UNIVERSIDAD PABLO DE OLAVIDE
Publications
2026
D. Martín-Pérez and F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez
Hybrid Classical-Quantum Transfer Learning with Noisy Quantum Circuits Journal Article
In: Computer Modeling in Engineering and Sciences, vol. 147, pp. 82712, 2026.
Abstract | Links | BibTeX | Tags: quantum computing, transfer learning
@article{MARTIN26_CMES,
title = {Hybrid Classical-Quantum Transfer Learning with Noisy Quantum Circuits},
author = {D. Martín-Pérez and F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez},
url = {https://www.techscience.com/CMES/online/detail/27506},
doi = {http://doi.org/10.32604/cmes.2026.082712},
year = {2026},
date = {2026-07-09},
urldate = {2026-07-01},
journal = {Computer Modeling in Engineering and Sciences},
volume = {147},
pages = {82712},
publisher = {Tech Science Press},
abstract = {Hybrid classical-quantum architectures have emerged as a practical response to the data and compute demands of modern deep learning, since a pretrained classical backbone can carry the feature extraction while a compact quantum head provides the trainable component. Quantum transfer learning is the most active instance of this idea. However, existing quantum transfer learning pipelines have been evaluated in isolation, typically on a single software framework and without a structured treatment of noise or statistical significance, which makes it difficult to assess how this paradigm contributes over fair classical baselines. A methodological benchmark for quantum transfer learning is proposed in this paper. The benchmark couples a common set of frozen pretrained convolutional backbones to several classical and quantum classification heads, implemented in both PennyLane and Qiskit, so that the contribution of nonlinearity, software framework, and quantum component can be analyzed separately. The benchmark has been applied to heterogeneous image datasets covering medical, biological, industrial, and general-vision domains, and most configurations have been executed in three environments: ideal simulation, noisy emulation calibrated on IBM Heron r2 and real IBM quantum hardware. Complementary analyses isolate the contribution of individual noise channels, examine scalability with qubit count and circuit depth and assess the presence of barren plateaus. Statistical reliability is ensured through multiple random seeds and pairwise Wilcoxon signed-rank tests, providing the first controlled cross-framework assessment of quantum transfer learning under calibrated noise and on real hardware.},
keywords = {quantum computing, transfer learning},
pubstate = {published},
tppubtype = {article}
}
Hybrid classical-quantum architectures have emerged as a practical response to the data and compute demands of modern deep learning, since a pretrained classical backbone can carry the feature extraction while a compact quantum head provides the trainable component. Quantum transfer learning is the most active instance of this idea. However, existing quantum transfer learning pipelines have been evaluated in isolation, typically on a single software framework and without a structured treatment of noise or statistical significance, which makes it difficult to assess how this paradigm contributes over fair classical baselines. A methodological benchmark for quantum transfer learning is proposed in this paper. The benchmark couples a common set of frozen pretrained convolutional backbones to several classical and quantum classification heads, implemented in both PennyLane and Qiskit, so that the contribution of nonlinearity, software framework, and quantum component can be analyzed separately. The benchmark has been applied to heterogeneous image datasets covering medical, biological, industrial, and general-vision domains, and most configurations have been executed in three environments: ideal simulation, noisy emulation calibrated on IBM Heron r2 and real IBM quantum hardware. Complementary analyses isolate the contribution of individual noise channels, examine scalability with qubit count and circuit depth and assess the presence of barren plateaus. Statistical reliability is ensured through multiple random seeds and pairwise Wilcoxon signed-rank tests, providing the first controlled cross-framework assessment of quantum transfer learning under calibrated noise and on real hardware.
D. Martín-Pérez and F. Rodríguez-Díaz and A. Troncoso and F. Martínez-Álvarez
SOCO 21st Conference on Computing Systems and Applications, Communications in Computer and Information Science 2026.
Abstract | Links | BibTeX | Tags: quantum computing
@conference{SOCO26_Martin,
title = {Mitigating Catastrophic Forgetting in Quantum Neural Networks via Hybrid Topologies and Dark Experience Replay},
author = {D. Martín-Pérez and F. Rodríguez-Díaz and A. Troncoso and F. Martínez-Álvarez},
url = {https://link.springer.com/chapter/10.1007/978-3-032-29254-4_37},
doi = {https://doi.org/10.1007/978-3-032-29254-4_37},
year = {2026},
date = {2026-06-16},
booktitle = {SOCO 21st Conference on Computing Systems and Applications},
pages = {463–472},
series = {Communications in Computer and Information Science},
abstract = {Quantum Machine Learning has shown significant promise but suffers from catastrophic forgetting in continual learning scenarios. In this work, a hybrid classical-quantum approach to mitigate catastrophic forgetting is proposed, based on a rigorous evaluation of how the topology of Variational Quantum Circuits interacts with Dark Experience Replay. By constructing a controlled Noisy Intermediate-Scale Quantum simulation environment based on the Fashion-MNIST dataset, the forgetting dynamics are analyzed across three distinct quantum circuit ansatzes: Highly Expressive, Hardware-Efficient and Tree Tensor Networks. Extensive evaluation demonstrates that, when relying on naive fine-tuning, dense quantum entanglement structures exhibit more than 30% catastrophic forgetting across sequential tasks. However, by using Dark Experience Replay specifically, the hybrid network effectively anchors past distributions via logit distillation, reducing catastrophic forgetting margins to under 5% and robustly retaining previously acquired knowledge. This study provides an empirical foundation for deploying scalable lifelong learning on Noisy Intermediate-Scale Quantum devices using state-of-the-art memory buffers.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {conference}
}
Quantum Machine Learning has shown significant promise but suffers from catastrophic forgetting in continual learning scenarios. In this work, a hybrid classical-quantum approach to mitigate catastrophic forgetting is proposed, based on a rigorous evaluation of how the topology of Variational Quantum Circuits interacts with Dark Experience Replay. By constructing a controlled Noisy Intermediate-Scale Quantum simulation environment based on the Fashion-MNIST dataset, the forgetting dynamics are analyzed across three distinct quantum circuit ansatzes: Highly Expressive, Hardware-Efficient and Tree Tensor Networks. Extensive evaluation demonstrates that, when relying on naive fine-tuning, dense quantum entanglement structures exhibit more than 30% catastrophic forgetting across sequential tasks. However, by using Dark Experience Replay specifically, the hybrid network effectively anchors past distributions via logit distillation, reducing catastrophic forgetting margins to under 5% and robustly retaining previously acquired knowledge. This study provides an empirical foundation for deploying scalable lifelong learning on Noisy Intermediate-Scale Quantum devices using state-of-the-art memory buffers.
F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez
Gate Value Quantification for Efficient Quantum Machine Learning Circuits Conference
SOCO 21st Conference on Computing Systems and Applications, Communications in Computer and Information Science 2026.
Abstract | Links | BibTeX | Tags: quantum computing
@conference{SOCO26_Rodriguez,
title = {Gate Value Quantification for Efficient Quantum Machine Learning Circuits},
author = {F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez},
url = {https://link.springer.com/chapter/10.1007/978-3-032-29254-4_34},
doi = {https://doi.org/10.1007/978-3-032-29254-4_34},
year = {2026},
date = {2026-06-16},
booktitle = {SOCO 21st Conference on Computing Systems and Applications},
pages = {429–438},
series = {Communications in Computer and Information Science},
abstract = {Quantum machine learning circuits are limited by depth, the number of two-qubit gates, and execution cost, which motivates the elimination of operations that contribute little to predictive performance. This paper presents a gate value methodology that quantifies the contribution of individual gates and iteratively derives reduced circuit configurations to improve the trade-off between accuracy and time. Gate relevance is calculated using two complementary strategies: a fidelity-based strategy and an entanglement-based strategy. Experimental configurations are ranked by best accuracy and best time. Finally, experiments with several binary classification datasets in an ideal simulation environment show that moderate gate removal can preserve, and sometimes improve, accuracy while reducing execution time.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {conference}
}
Quantum machine learning circuits are limited by depth, the number of two-qubit gates, and execution cost, which motivates the elimination of operations that contribute little to predictive performance. This paper presents a gate value methodology that quantifies the contribution of individual gates and iteratively derives reduced circuit configurations to improve the trade-off between accuracy and time. Gate relevance is calculated using two complementary strategies: a fidelity-based strategy and an entanglement-based strategy. Experimental configurations are ranked by best accuracy and best time. Finally, experiments with several binary classification datasets in an ideal simulation environment show that moderate gate removal can preserve, and sometimes improve, accuracy while reducing execution time.
E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez
Communications in Computer and Information Science, Springer, vol. 3046, 2026, ISBN: 978-3-032-29254-4.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{SOCO2026_proceedings,
title = {Proceedings of the 21st International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2026) Marbella, Spain, June 18–19, 2026},
author = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez},
editor = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez},
url = {https://link.springer.com/book/10.1007/978-3-032-29254-4},
doi = {https://doi.org/10.1007/978-3-032-29254-4},
isbn = {978-3-032-29254-4},
year = {2026},
date = {2026-06-14},
urldate = {2026-06-14},
volume = {3046},
publisher = {Communications in Computer and Information Science, Springer},
series = {Communications in Computer and Information Science},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe
Lecture Notes in Artificial Intelligence, Springer, vol. 16596, 2026, ISBN: 978-3-032-29292-6.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{HAIS2026,
title = {Proceedings of the 21st International Conference on Hybrid Artificial Intelligent Systems (HAIS 2026) Marbella, Spain, June 18-19, 2026},
author = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
editor = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
url = {https://link.springer.com/book/10.1007/978-3-032-29292-6},
doi = {https://doi.org/10.1007/978-3-032-29292-6},
isbn = {978-3-032-29292-6},
year = {2026},
date = {2026-06-13},
urldate = {2025-06-13},
volume = {16596},
publisher = {Lecture Notes in Artificial Intelligence, Springer},
series = {Lecture Notes in Artificial Intelligence},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
H. Affou and F. Rodríguez-Díaz and F. Martínez-Álvarez and J. M. López-Guede
IWINAC 11th International Work-Conference on the Interplay Between Natural and Artificial Computation, Lecture Notes in Computer Science 2026.
Abstract | Links | BibTeX | Tags: quantum computing
@conference{IWINAC26_Affou,
title = {Quantum–Classical Hybrid Genetic Evolutionary Algorithm for Traffic Signal Timing Optimization: A Case Study in the City of Vitoria-Gasteiz},
author = {H. Affou and F. Rodríguez-Díaz and F. Martínez-Álvarez and J. M. López-Guede},
url = {https://link.springer.com/chapter/10.1007/978-3-032-27317-8_26},
doi = {https://doi.org/10.1007/978-3-032-27317-8_26},
year = {2026},
date = {2026-05-30},
booktitle = {IWINAC 11th International Work-Conference on the Interplay Between Natural and Artificial Computation},
pages = {268–277},
series = {Lecture Notes in Computer Science},
abstract = {Intelligent traffic signal control has become a central research topic in urban mobility, aiming to reduce congestion and improve operational efficiency under increasingly complex traffic conditions. Although classical evolutionary algorithms have shown strong performance in traffic signal optimization, the practical integration of quantum computing into evolutionary processes remains largely unexplored, particularly under realistic microscopic traffic simulation. This paper presents a novel quantum genetic evolutionary approach for traffic signal timing optimization, implemented in a real signalized roundabout in Vitoria-Gasteiz, Spain, using the Simulation of Urban Mobility. The proposed method combines a classical GA for global exploration with a Grover-inspired quantum module embedded directly within the evolutionary cycle. Unlike conventional hybrid schemes in which quantum routines are treated as external solvers or purely simulated components, the proposed quantum GA is executed on real IBM Quantum hardware under NISQ conditions. Comparative experiments against fixed time control and a purely classical GA demonstrate substantial reductions in delay and improved convergence stability.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {conference}
}
Intelligent traffic signal control has become a central research topic in urban mobility, aiming to reduce congestion and improve operational efficiency under increasingly complex traffic conditions. Although classical evolutionary algorithms have shown strong performance in traffic signal optimization, the practical integration of quantum computing into evolutionary processes remains largely unexplored, particularly under realistic microscopic traffic simulation. This paper presents a novel quantum genetic evolutionary approach for traffic signal timing optimization, implemented in a real signalized roundabout in Vitoria-Gasteiz, Spain, using the Simulation of Urban Mobility. The proposed method combines a classical GA for global exploration with a Grover-inspired quantum module embedded directly within the evolutionary cycle. Unlike conventional hybrid schemes in which quantum routines are treated as external solvers or purely simulated components, the proposed quantum GA is executed on real IBM Quantum hardware under NISQ conditions. Comparative experiments against fixed time control and a purely classical GA demonstrate substantial reductions in delay and improved convergence stability.
N. Ullah and F. Martínez-Álvarez and I. De Falco and G. Sannino
An explainable deep learning method based on spectral co-clustering for ozone time series forecasting Journal Article
In: Applied Intelligence, vol. 56, pp. 267, 2026.
Abstract | Links | BibTeX | Tags: deep learning, time series, XAI
@article{ULLAH26_APIN,
title = {An explainable deep learning method based on spectral co-clustering for ozone time series forecasting},
author = {N. Ullah and F. Martínez-Álvarez and I. De Falco and G. Sannino},
url = {https://link.springer.com/article/10.1007/s10489-026-07278-y},
doi = {https://doi.org/10.1007/s10489-026-07278-y},
year = {2026},
date = {2026-05-21},
urldate = {2026-05-21},
journal = {Applied Intelligence},
volume = {56},
pages = {267},
publisher = {Springer},
abstract = {Tropospheric ozone forecasting is critical for public health, yet the deep learning models that achieve high accuracy often function as black boxes. This lack of transparency, along with the inability of popular explainability techniques like SHapley Additive exPlanations (SHAP) to capture essential temporal dependencies, limits their practical utility and trustworthiness in environmental management. To address this, we propose a novel framework, eXplainable Deep Learning with Spectral Co-clustering for Time Series, that integrates spectral co-clustering to enhance forecasting performance and provide post-hoc structured interpretability for air-quality time series. The methodology comprises data preprocessing, feature engineering (including lagging, rolling statistics, and time-based features), and deep learning architectures (Multilayer Perceptron, Gated Recurrent Unit, and hybrid models). Bayesian optimization is used to fine-tune hyperparameters. The core contribution is a spectral co-clustering technique that simultaneously partitions features and time instances into co-clusters, revealing critical inter-feature relationships and temporal patterns that drive predictions. The framework was rigorously validated through extensive experiments on data from five air quality monitoring stations. The proposed approach achieved RMSE values ranging from 0.73 to 6.08, significantly outperforming existing methods, including a temporal LSTM baseline, with performance improvements of approximately 59.19% to 95.65%. Results demonstrate that the proposed approach not only achieves high forecasting accuracy but also, through post-hoc heatmap visualizations of the objectively selected best-performing co-cluster, identifies the key features and time periods governing model predictions, thereby offering an interpretable understanding of the temporal and feature-level drivers associated with ozone variability. Thus, a transparent and effective solution for ozone forecasting is proposed, with a modular design generalizable to other environmental time series prediction tasks.},
keywords = {deep learning, time series, XAI},
pubstate = {published},
tppubtype = {article}
}
Tropospheric ozone forecasting is critical for public health, yet the deep learning models that achieve high accuracy often function as black boxes. This lack of transparency, along with the inability of popular explainability techniques like SHapley Additive exPlanations (SHAP) to capture essential temporal dependencies, limits their practical utility and trustworthiness in environmental management. To address this, we propose a novel framework, eXplainable Deep Learning with Spectral Co-clustering for Time Series, that integrates spectral co-clustering to enhance forecasting performance and provide post-hoc structured interpretability for air-quality time series. The methodology comprises data preprocessing, feature engineering (including lagging, rolling statistics, and time-based features), and deep learning architectures (Multilayer Perceptron, Gated Recurrent Unit, and hybrid models). Bayesian optimization is used to fine-tune hyperparameters. The core contribution is a spectral co-clustering technique that simultaneously partitions features and time instances into co-clusters, revealing critical inter-feature relationships and temporal patterns that drive predictions. The framework was rigorously validated through extensive experiments on data from five air quality monitoring stations. The proposed approach achieved RMSE values ranging from 0.73 to 6.08, significantly outperforming existing methods, including a temporal LSTM baseline, with performance improvements of approximately 59.19% to 95.65%. Results demonstrate that the proposed approach not only achieves high forecasting accuracy but also, through post-hoc heatmap visualizations of the objectively selected best-performing co-cluster, identifies the key features and time periods governing model predictions, thereby offering an interpretable understanding of the temporal and feature-level drivers associated with ozone variability. Thus, a transparent and effective solution for ozone forecasting is proposed, with a modular design generalizable to other environmental time series prediction tasks.
A. Moayedikia and A. Troncoso
Bridging training and merging through momentum-aware optimization Journal Article
In: Information Sciences, vol. 745, pp. 123402, 2026.
Abstract | Links | BibTeX | Tags: deep learning, large language model
@article{Troncoso2026,
title = {Bridging training and merging through momentum-aware optimization},
author = {A. Moayedikia and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S0020025526003336},
doi = {10.1016/j.ins.2026.123402},
year = {2026},
date = {2026-05-01},
urldate = {2026-05-01},
journal = {Information Sciences},
volume = {745},
pages = {123402},
abstract = {Training large neural networks and merging task-specific models both exploit low-rank structure and require parameter importance estimation, yet these challenges have been pursued in isolation. Current workflows compute curvature information during training, discard it, then recompute similar information for merging—wasting computation and discarding valuable trajectory data. We introduce a unified framework that maintains factorized momentum and curvature statistics during training, then reuses this information for geometry-aware model composition. The proposed method incurs modest memory overhead (approximately 30% over AdamW) to accumulate task saliency scores that enable curvature-aware merging. These scores, computed as a byproduct of optimization, provide importance estimates comparable to post-hoc Fisher computation while producing merge-ready models directly from training. We establish convergence guarantees for non-convex objectives with approximation error bounded by gradient singular value decay. On natural language understanding benchmarks, curvature-aware parameter selection outperforms magnitude-only baselines across all sparsity levels, with multi-task merging improving 1.6% over strong baselines. The proposed framework exhibits rank-invariant convergence and superior hyperparameter robustness compared to existing low-rank optimizers. By treating the optimization trajectory as a reusable asset rather than discarding it, our approach demonstrates that training-time curvature information suffices for effective model composition, enabling a unified training-merging pipeline.},
keywords = {deep learning, large language model},
pubstate = {published},
tppubtype = {article}
}
Training large neural networks and merging task-specific models both exploit low-rank structure and require parameter importance estimation, yet these challenges have been pursued in isolation. Current workflows compute curvature information during training, discard it, then recompute similar information for merging—wasting computation and discarding valuable trajectory data. We introduce a unified framework that maintains factorized momentum and curvature statistics during training, then reuses this information for geometry-aware model composition. The proposed method incurs modest memory overhead (approximately 30% over AdamW) to accumulate task saliency scores that enable curvature-aware merging. These scores, computed as a byproduct of optimization, provide importance estimates comparable to post-hoc Fisher computation while producing merge-ready models directly from training. We establish convergence guarantees for non-convex objectives with approximation error bounded by gradient singular value decay. On natural language understanding benchmarks, curvature-aware parameter selection outperforms magnitude-only baselines across all sparsity levels, with multi-task merging improving 1.6% over strong baselines. The proposed framework exhibits rank-invariant convergence and superior hyperparameter robustness compared to existing low-rank optimizers. By treating the optimization trajectory as a reusable asset rather than discarding it, our approach demonstrates that training-time curvature information suffices for effective model composition, enabling a unified training-merging pipeline.
F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez
A Survey of Quantum Machine Learning: Foundations, Algorithms, Frameworks, Data and Applications Journal Article
In: ACM Computing Surveys, vol. 58, iss. 4, pp. 1-35, 2026.
Abstract | Links | BibTeX | Tags: quantum computing
@article{CSUR2025,
title = {A Survey of Quantum Machine Learning: Foundations, Algorithms, Frameworks, Data and Applications},
author = {F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez},
url = {https://dl.acm.org/doi/10.1145/3764582},
doi = {10.1145/3764582},
year = {2026},
date = {2026-03-02},
urldate = {2026-03-02},
journal = {ACM Computing Surveys},
volume = {58},
issue = {4},
pages = {1-35},
abstract = {Quantum machine learning combines quantum computing with machine learning to solve complex computational problems more efficiently than classical approaches. This survey provides an introduction to the foundations, algorithms, frameworks, data and applications of quantum machine learning, serving as a resource for researchers and practitioners. We begin by reviewing existing surveys to identify gaps that this work addresses, followed by a detailed discussion of the foundational principles of quantum mechanics and machine learning essential for quantum machine learning. Key algorithms are examined, highlighting their mechanisms, advantages, and applications across various domains. Current frameworks and platforms for implementing quantum machine learning algorithms are explored, emphasizing their unique features and suitability for different contexts. Existing quantum datasets for practical usage are also reported and commented on. This survey also reviews over 135 articles, categorized into theoretical and practical contributions, to identify key advances, limitations, and application areas within quantum machine learning. Critical challenges such as hardware limitations, error rates, and scalability are analyzed to detect the obstacles that must be addressed for practical deployment. By synthesizing these elements into a structured overview, this survey aims at serving as both an introduction and a guide for advancing research and development in this disruptive field.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {article}
}
Quantum machine learning combines quantum computing with machine learning to solve complex computational problems more efficiently than classical approaches. This survey provides an introduction to the foundations, algorithms, frameworks, data and applications of quantum machine learning, serving as a resource for researchers and practitioners. We begin by reviewing existing surveys to identify gaps that this work addresses, followed by a detailed discussion of the foundational principles of quantum mechanics and machine learning essential for quantum machine learning. Key algorithms are examined, highlighting their mechanisms, advantages, and applications across various domains. Current frameworks and platforms for implementing quantum machine learning algorithms are explored, emphasizing their unique features and suitability for different contexts. Existing quantum datasets for practical usage are also reported and commented on. This survey also reviews over 135 articles, categorized into theoretical and practical contributions, to identify key advances, limitations, and application areas within quantum machine learning. Critical challenges such as hardware limitations, error rates, and scalability are analyzed to detect the obstacles that must be addressed for practical deployment. By synthesizing these elements into a structured overview, this survey aims at serving as both an introduction and a guide for advancing research and development in this disruptive field.
A. Vellinger and F. Rodríguez-Díaz and F. Divina and J. F. Torres
Forecasting livestock activity through interpretable neuroevolutionary transfer learning Journal Article
In: Logic Journal of the IGPL, vol. 34, pp. jzaf034, 2026.
Abstract | Links | BibTeX | Tags: deep learning, forecasting, time series, XAI
@article{VELLINGER26,
title = {Forecasting livestock activity through interpretable neuroevolutionary transfer learning},
author = {A. Vellinger and F. Rodríguez-Díaz and F. Divina and J. F. Torres},
url = {https://academic.oup.com/jigpal/article/34/1/jzaf034/8487859},
doi = {10.1093/jigpal/jzaf034},
year = {2026},
date = {2026-02-16},
urldate = {2026-02-16},
journal = {Logic Journal of the IGPL},
volume = {34},
pages = {jzaf034},
publisher = {Oxford academic},
abstract = {In this paper, we describe a neuroevolutionary approach to livestock activity forecasting, specifically targeting the prediction of Iberian pigs movements. We successfully integrated Transfer Learning to save computational time and used an Explainable Artificial Intelligence technique to provide valuable insights from the model predictions. Inspired by previous work, we employ Deep Evolutionary Network Structured Representation to optimize both Long Short-Term Memory networks and Convolutional Neural Networks using genetic algorithms and dynamic structured grammatical evolution, and we compare the results with other commonly used approaches for time series forecasting. Experimental results demonstrate the superior performance of the proposed Long Short-Term Memory models over more traditional methods, highlighting their precision and consistency in predicting livestock activities. Furthermore, the application of Explainable Artificial Intelligence techniques enable to gain a deeper understanding and trust in AI-driven decisions within precision livestock farming.},
keywords = {deep learning, forecasting, time series, XAI},
pubstate = {published},
tppubtype = {article}
}
In this paper, we describe a neuroevolutionary approach to livestock activity forecasting, specifically targeting the prediction of Iberian pigs movements. We successfully integrated Transfer Learning to save computational time and used an Explainable Artificial Intelligence technique to provide valuable insights from the model predictions. Inspired by previous work, we employ Deep Evolutionary Network Structured Representation to optimize both Long Short-Term Memory networks and Convolutional Neural Networks using genetic algorithms and dynamic structured grammatical evolution, and we compare the results with other commonly used approaches for time series forecasting. Experimental results demonstrate the superior performance of the proposed Long Short-Term Memory models over more traditional methods, highlighting their precision and consistency in predicting livestock activities. Furthermore, the application of Explainable Artificial Intelligence techniques enable to gain a deeper understanding and trust in AI-driven decisions within precision livestock farming.
A. Gil-Gamboa and J. E. Sánchez-López and J. Solís-García and M. A. Delgado and I. Román and J. Franquelo and J. C. Riquelme and A. Troncoso
Enhancing energy availability based on explainable two-level clustering for root-cause diagnosis of production loss events in wind farms Journal Article
In: Renewable Energy, vol. 267, pp. 125656, 2026.
Abstract | Links | BibTeX | Tags: clustering, energy, feature selection, forecasting, pattern recognition, time series
@article{GILGAMBOA2026125656,
title = {Enhancing energy availability based on explainable two-level clustering for root-cause diagnosis of production loss events in wind farms},
author = { A. Gil-Gamboa and J. E. Sánchez-López and J. Solís-García and M. A. Delgado and I. Román and J. Franquelo and J. C. Riquelme and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S0960148126004817},
doi = {10.1016/j.renene.2026.125656},
year = {2026},
date = {2026-02-14},
urldate = {2026-02-14},
journal = {Renewable Energy},
volume = {267},
pages = {125656},
publisher = {Elsevier},
abstract = {The economic viability of wind energy is fundamentally linked to the maximization of availability and the minimization of production losses caused by unscheduled downtime. Therefore, a quick and accurate diagnosis of the root cause of failures is critical to restoring power generation efficiency. It is vital to analyze what caused failures in the operation of wind farms so that the necessary actions can be taken to get the wind turbine producing power again in the shortest possible time. In this paper, we propose to discover patterns that can help diagnose the root cause of a failure in one or several wind turbines. For this purpose, a specific definition of an incident as an energy production loss event is proposed, as well as a selection of variables that characterize such incidents. Addressing the limitations of supervised learning in label-scarce environments, an explainable methodology based on unsupervised learning is carried out at two levels: a first level, to determine different groups of incidents through clustering, and a second level to obtain groups of incidents sharing alarms, called subclusters. Finally, a rule summarizing each subcluster is provided to explain the patterns that identify each root cause, offering rule-based explainability that deep learning models lack. Two real-world datasets composed of sensor signal records together with alarms that occurred at wind farms located in Spain and Argentina were used to evaluate the model. Experimental results show a high diagnostic accuracy, which directly facilitates a reduction in average repair time and an improvement in annual energy production.},
keywords = {clustering, energy, feature selection, forecasting, pattern recognition, time series},
pubstate = {published},
tppubtype = {article}
}
The economic viability of wind energy is fundamentally linked to the maximization of availability and the minimization of production losses caused by unscheduled downtime. Therefore, a quick and accurate diagnosis of the root cause of failures is critical to restoring power generation efficiency. It is vital to analyze what caused failures in the operation of wind farms so that the necessary actions can be taken to get the wind turbine producing power again in the shortest possible time. In this paper, we propose to discover patterns that can help diagnose the root cause of a failure in one or several wind turbines. For this purpose, a specific definition of an incident as an energy production loss event is proposed, as well as a selection of variables that characterize such incidents. Addressing the limitations of supervised learning in label-scarce environments, an explainable methodology based on unsupervised learning is carried out at two levels: a first level, to determine different groups of incidents through clustering, and a second level to obtain groups of incidents sharing alarms, called subclusters. Finally, a rule summarizing each subcluster is provided to explain the patterns that identify each root cause, offering rule-based explainability that deep learning models lack. Two real-world datasets composed of sensor signal records together with alarms that occurred at wind farms located in Spain and Argentina were used to evaluate the model. Experimental results show a high diagnostic accuracy, which directly facilitates a reduction in average repair time and an improvement in annual energy production.
N. Ullah and A. M. Chacón-Maldonado and F. Martínez-Álvarez and I. De Falco and G. Sannino
A Novel Knowledge Distillation and Hybrid Explainability Approach for Phenology Stage Classification from Multi-Source Time Series Journal Article
In: Information Fusion, vol. 131, pp. 104158, 2026.
Abstract | Links | BibTeX | Tags: association rules, precision agriculture, XAI
@article{ullah2026novel,
title = {A Novel Knowledge Distillation and Hybrid Explainability Approach for Phenology Stage Classification from Multi-Source Time Series},
author = { N. Ullah and A. M. Chacón-Maldonado and F. Martínez-Álvarez and I. De Falco and G. Sannino},
url = {https://www.sciencedirect.com/science/article/pii/S1566253526000370},
doi = {10.1016/j.inffus.2026.104158},
year = {2026},
date = {2026-01-16},
urldate = {2026-01-16},
journal = {Information Fusion},
volume = {131},
pages = {104158},
publisher = {Elsevier},
abstract = {Accurate phenological stage classification is crucial for addressing global challenges to food security posed by climate change, water scarcity, and land degradation. It enables precision agriculture by optimizing key interventions such as irrigation, fertilization, and pest control. While deep learning offers powerful tools, existing methods face four key limitations: reliance on narrow features and models, limited long-term forecasting capability, computational inefficiency, and opaque, unvalidated explanations. To overcome these limitations, this paper presents a deep learning framework for phenology classification, utilizing multi-source time series data from satellite imagery, meteorological stations, and field observations. The approach emphasizes temporal consistency, spatial adaptability, computational efficiency, and explainability. A feature engineering pipeline extracts temporal dynamics via lag features, rolling statistics, Fourier transforms and seasonal encodings. Feature selection combines incremental strategies with classical filter, wrapper, and embedded methods. Deep learning models across multiple paradigms-feedforward, recurrent, convolutional, and attention-based-are benchmarked under multi-horizon forecasting tasks. To reduce model complexity while preserving performance where possible, the framework employs knowledge distillation, transferring predictive knowledge from complex teacher models to compact and deployable student models. For model interpretability, a new Hybrid SHAP-Association Rule Explainability approach is proposed, integrating model-driven and data-driven explanations. Agreement between views is quantified using trust metrics: precision@k, coverage, and Jaccard similarity, with a retraining-based validation mechanism. Experiments on phenology data from Andalusia demonstrate high accuracy, strong generalizability, trustworthy explanations and resource-efficient phenology monitoring in agricultural systems.},
keywords = {association rules, precision agriculture, XAI},
pubstate = {published},
tppubtype = {article}
}
Accurate phenological stage classification is crucial for addressing global challenges to food security posed by climate change, water scarcity, and land degradation. It enables precision agriculture by optimizing key interventions such as irrigation, fertilization, and pest control. While deep learning offers powerful tools, existing methods face four key limitations: reliance on narrow features and models, limited long-term forecasting capability, computational inefficiency, and opaque, unvalidated explanations. To overcome these limitations, this paper presents a deep learning framework for phenology classification, utilizing multi-source time series data from satellite imagery, meteorological stations, and field observations. The approach emphasizes temporal consistency, spatial adaptability, computational efficiency, and explainability. A feature engineering pipeline extracts temporal dynamics via lag features, rolling statistics, Fourier transforms and seasonal encodings. Feature selection combines incremental strategies with classical filter, wrapper, and embedded methods. Deep learning models across multiple paradigms-feedforward, recurrent, convolutional, and attention-based-are benchmarked under multi-horizon forecasting tasks. To reduce model complexity while preserving performance where possible, the framework employs knowledge distillation, transferring predictive knowledge from complex teacher models to compact and deployable student models. For model interpretability, a new Hybrid SHAP-Association Rule Explainability approach is proposed, integrating model-driven and data-driven explanations. Agreement between views is quantified using trust metrics: precision@k, coverage, and Jaccard similarity, with a retraining-based validation mechanism. Experiments on phenology data from Andalusia demonstrate high accuracy, strong generalizability, trustworthy explanations and resource-efficient phenology monitoring in agricultural systems.
P. Reina-Jiménez and M.J. Jiménez-Navarro and G. Asencio-Cortés and F. Martínez-Álvarez and M. Martínez-Ballesteros
A novel interpretable ozone forecasting approach based on deep learning with masked residual connections Journal Article
In: Environmental Modelling and Software, vol. 198, pp. 106878, 2026.
Abstract | Links | BibTeX | Tags: deep learning, forecasting, time series, XAI
@article{REINAJIMENEZ2026,
title = {A novel interpretable ozone forecasting approach based on deep learning with masked residual connections},
author = {P. Reina-Jiménez and M.J. Jiménez-Navarro and G. Asencio-Cortés and F. Martínez-Álvarez and M. Martínez-Ballesteros},
url = {https://www.sciencedirect.com/science/article/pii/S1364815226000253},
doi = {https://doi.org/10.1016/j.envsoft.2026.106878},
year = {2026},
date = {2026-01-01},
urldate = {2026-01-01},
journal = {Environmental Modelling and Software},
volume = {198},
pages = {106878},
abstract = {Air pollution is a growing threat, especially in low- and middle-income countries, causing over 4 million premature deaths annually. Ground-level ozone is a major concern, demanding accurate and interpretable prediction systems for effective public health management. However, existing time-series forecasting methods struggle to capture both linear and nonlinear dependencies in atmospheric data. This study introduces ResSelNet, a novel Residual Selection Network that integrates masked residual connections and embedded feature selection within a unified deep learning architecture. The model dynamically determines the optimal processing depth for each feature, allowing linear relationships to bypass nonlinear transformations while capturing complex patterns when necessary. Applied to five monitoring stations across Andalusia (Spain), ResSelNet consistently outperformed state-of-the-art baselines, achieving 8%–12% lower RMSE and MAE than LSTM and Transformer models. Beyond accuracy, the framework improves interpretability and robustness, revealing the hierarchical relevance of meteorological and pollutant variables. ResSelNet therefore offers an effective and explainable solution for multi-horizon environmental time-series forecasting.},
keywords = {deep learning, forecasting, time series, XAI},
pubstate = {published},
tppubtype = {article}
}
Air pollution is a growing threat, especially in low- and middle-income countries, causing over 4 million premature deaths annually. Ground-level ozone is a major concern, demanding accurate and interpretable prediction systems for effective public health management. However, existing time-series forecasting methods struggle to capture both linear and nonlinear dependencies in atmospheric data. This study introduces ResSelNet, a novel Residual Selection Network that integrates masked residual connections and embedded feature selection within a unified deep learning architecture. The model dynamically determines the optimal processing depth for each feature, allowing linear relationships to bypass nonlinear transformations while capturing complex patterns when necessary. Applied to five monitoring stations across Andalusia (Spain), ResSelNet consistently outperformed state-of-the-art baselines, achieving 8%–12% lower RMSE and MAE than LSTM and Transformer models. Beyond accuracy, the framework improves interpretability and robustness, revealing the hierarchical relevance of meteorological and pollutant variables. ResSelNet therefore offers an effective and explainable solution for multi-horizon environmental time-series forecasting.
2025
M. Solís and A. Gil-Gamboa and A. Troncoso
Metalearning for improving time series forecasting based on deep learning: A water case study Journal Article
In: Results in Engineering, vol. 28, pp. 107541, 2025.
Links | BibTeX | Tags: deep learning, forecasting, time series
@article{RING2025_Martin,
title = {Metalearning for improving time series forecasting based on deep learning: A water case study},
author = {M. Solís and A. Gil-Gamboa and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S2590123025035960},
doi = {10.1016/j.rineng.2025.107541},
year = {2025},
date = {2025-12-09},
urldate = {2025-12-09},
journal = {Results in Engineering},
volume = {28},
pages = {107541},
keywords = {deep learning, forecasting, time series},
pubstate = {published},
tppubtype = {article}
}
A. M. Chacón-Maldonado and A. R. Troncoso-García and G. Asencio-Cortés and A. Troncoso
Improving monsoon forecasting based on feature selection and explainable artificial intelligence Journal Article
In: Applied Soft Computing, vol. 185, pp. 114053, 2025.
Links | BibTeX | Tags: feature selection, natural disasters, XAI
@article{ASOC2024,
title = {Improving monsoon forecasting based on feature selection and explainable artificial intelligence},
author = {A. M. Chacón-Maldonado and A. R. Troncoso-García and G. Asencio-Cortés and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S1568494625013663},
doi = {10.1016/j.asoc.2025.114053},
year = {2025},
date = {2025-12-02},
urldate = {2025-12-02},
journal = {Applied Soft Computing},
volume = {185},
pages = {114053},
keywords = {feature selection, natural disasters, XAI},
pubstate = {published},
tppubtype = {article}
}
A. R. Troncoso-García and M. Martínez-Ballesteros and F. Martínez-Álvarez and A. Troncoso
Feature Importance in Association Rule-Based Explanations for Time Series Forecasting Conference
IDEAL 26th International Conference on Intelligent Data Engineering and Automated Learning, Lecture Notes in Artificial Intelligence 2025.
Links | BibTeX | Tags: association rules, forecasting, time series, XAI
@conference{IDEAL2025_Angela,
title = {Feature Importance in Association Rule-Based Explanations for Time Series Forecasting},
author = {A. R. Troncoso-García and M. Martínez-Ballesteros and F. Martínez-Álvarez and A. Troncoso},
url = {https://link.springer.com/chapter/10.1007/978-3-032-10489-2_20},
doi = {10.1007/978-3-032-10489-2_20},
year = {2025},
date = {2025-11-13},
urldate = {2025-11-13},
booktitle = {IDEAL 26th International Conference on Intelligent Data Engineering and Automated Learning},
series = {Lecture Notes in Artificial Intelligence},
keywords = {association rules, forecasting, time series, XAI},
pubstate = {published},
tppubtype = {conference}
}
A. M. Chacón-Maldonado and N. Martínez Van der Looven and G. Asencio-Cortés and A. Troncoso
A New Transformer-Based Hybrid Model to Forecast Olive Fruit Fly Using Multimodal Data Conference
HAIS 20th International Conference on Hybrid Artificial Intelligent Systems, Lecture Notes in Artificial Intelligence 2025.
Links | BibTeX | Tags: deep learning, precision agriculture
@conference{HAIS2025_Andres,
title = {A New Transformer-Based Hybrid Model to Forecast Olive Fruit Fly Using Multimodal Data},
author = {A. M. Chacón-Maldonado and N. Martínez Van der Looven and G. Asencio-Cortés and A. Troncoso},
url = {https://doi.org/},
doi = {10.1007/978-3-032-08465-1_3},
year = {2025},
date = {2025-10-15},
urldate = {2025-10-15},
booktitle = {HAIS 20th International Conference on Hybrid Artificial Intelligent Systems},
pages = {27-38},
series = {Lecture Notes in Artificial Intelligence },
keywords = {deep learning, precision agriculture},
pubstate = {published},
tppubtype = {conference}
}
E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe
Lecture Notes in Artificial Intelligence, Springer, vol. 16203, 2025, ISBN: 978-3-032-08462-0.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{HAIS2025_part2,
title = {Proceedings of the 20th International Conference on Hybrid Artificial Intelligent Systems (HAIS 2025) Salamanca, Spain, October 16-17, 2025, Part II},
author = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
editor = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
url = {https://link.springer.com/book/10.1007/978-3-032-08462-0},
doi = {https://doi.org/10.1007/978-3-032-08462-0},
isbn = {978-3-032-08462-0},
year = {2025},
date = {2025-10-13},
urldate = {2025-10-13},
volume = {16203},
publisher = {Lecture Notes in Artificial Intelligence, Springer},
series = {Lecture Notes in Artificial Intelligence},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe
Lecture Notes in Artificial Intelligence, Springer, vol. 16202, 2025, ISBN: 978-3-032-08464-4.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{HAIS2025_part1,
title = {Proceedings of the 20th International Conference on Hybrid Artificial Intelligent Systems (HAIS 2025) Salamanca, Spain, October 16-17, 2025, Part I},
author = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
editor = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci and R. S. Filipe},
url = {https://link.springer.com/book/10.1007/978-3-032-08465-1},
doi = {https://doi.org/10.1007/978-3-032-08465-1},
isbn = {978-3-032-08464-4},
year = {2025},
date = {2025-10-12},
urldate = {2025-10-12},
volume = {16202},
publisher = {Lecture Notes in Artificial Intelligence, Springer},
series = {Lecture Notes in Artificial Intelligence},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
Z. Wang and I. Koprinska and M. Martínez-Ballesteros and A. Troncoso and B. Jeffries
AIED 26th International Conference on Artificial Intelligence in Education, 2025.
Links | BibTeX | Tags: association rules, education
@conference{AIED2025,
title = {Comparison of Explainable Machine Learning Methods for Early Prediction of Student Performance in Programming Courses},
author = {Z. Wang and I. Koprinska and M. Martínez-Ballesteros and A. Troncoso and B. Jeffries },
url = {https://link.springer.com/chapter/10.1007/978-3-031-99264-3_20},
doi = {https://doi.org/10.1007/978-3-031-99264-3_20},
year = {2025},
date = {2025-07-24},
urldate = {2025-07-24},
booktitle = {AIED 26th International Conference on Artificial Intelligence in Education},
keywords = {association rules, education},
pubstate = {published},
tppubtype = {conference}
}
A. M. Chacón-Maldonado and G. Asencio-Cortés and A. Troncoso
A multimodal hybrid deep learning approach for pest forecasting using time series and satellite images Journal Article
In: Information Fusion, vol. 124, pp. 103350, 2025.
Links | BibTeX | Tags: deep learning, precision agriculture
@article{INFFUSChacon2025,
title = { A multimodal hybrid deep learning approach for pest forecasting using time series and satellite images},
author = {A. M. Chacón-Maldonado and G. Asencio-Cortés and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S1566253525004233},
doi = {10.1016/j.inffus.2025.103350},
year = {2025},
date = {2025-07-01},
urldate = {2025-07-01},
journal = {Information Fusion},
volume = {124},
pages = {103350},
keywords = {deep learning, precision agriculture},
pubstate = {published},
tppubtype = {article}
}
C. Herruzo-Lodeiro and F. Rodríguez-Díaz and A. Troncoso and M. Martínez-Ballesteros
SAC 40th ACM/SIGAPP Symposium on Applied Computing, 2025.
Links | BibTeX | Tags: association rules, pattern recognition
@conference{SAC2025,
title = {Bioinspired evolutionary metaheuristic based on COVID spread for discovering numerical association rules},
author = {C. Herruzo-Lodeiro and F. Rodríguez-Díaz and A. Troncoso and M. Martínez-Ballesteros},
doi = {10.1145/3672608.3707787},
year = {2025},
date = {2025-07-01},
urldate = {2025-07-01},
booktitle = {SAC 40th ACM/SIGAPP Symposium on Applied Computing},
pages = {138-144},
keywords = {association rules, pattern recognition},
pubstate = {published},
tppubtype = {conference}
}
D. Martín-Pérez and F. Rodríguez-Díaz and A. Troncoso and F. Martínez-Álvarez
Comparative Study of Hybrid Classical–Quantum Transfer Learning Conference
SOCO 20th Conference on Computing Systems and Applications, Communications in Computer and Information Science 2025.
Abstract | Links | BibTeX | Tags: quantum computing
@conference{SOCO25_Martin,
title = {Comparative Study of Hybrid Classical–Quantum Transfer Learning},
author = {D. Martín-Pérez and F. Rodríguez-Díaz and A. Troncoso and F. Martínez-Álvarez},
url = {https://link.springer.com/chapter/10.1007/978-3-032-19763-4_31},
doi = {https://doi.org/10.1007/978-3-032-19763-4_31},
year = {2025},
date = {2025-06-15},
urldate = {2026-06-15},
booktitle = {SOCO 20th Conference on Computing Systems and Applications},
pages = {337–346},
series = {Communications in Computer and Information Science},
abstract = {The combination of classical deep neural networks with quantum circuits has recently attracted attention as a promising paradigm for the current era of noisy intermediate-scale quantum (NISQ) technology. In this research, we perform a detailed comparative analysis of three classical Convolutional Neural Network architectures (ResNet18, VGG16, and MobileNetV2) combined with two distinct variational quantum circuits (VQC) acting as classifiers. To carefully assess the quantum contribution, we also introduce a purely classical baseline with an equivalent parameter budget. Each model is trained on the Hymenoptera dataset, and performance is checked based on validation accuracy and training time. Our findings show the important trade-offs between computational cost and model complexity, offering practical guidance for future hybrid classical-quantum applications and clarifying the performance benefits of different quantum circuit designs.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {conference}
}
The combination of classical deep neural networks with quantum circuits has recently attracted attention as a promising paradigm for the current era of noisy intermediate-scale quantum (NISQ) technology. In this research, we perform a detailed comparative analysis of three classical Convolutional Neural Network architectures (ResNet18, VGG16, and MobileNetV2) combined with two distinct variational quantum circuits (VQC) acting as classifiers. To carefully assess the quantum contribution, we also introduce a purely classical baseline with an equivalent parameter budget. Each model is trained on the Hymenoptera dataset, and performance is checked based on validation accuracy and training time. Our findings show the important trade-offs between computational cost and model complexity, offering practical guidance for future hybrid classical-quantum applications and clarifying the performance benefits of different quantum circuit designs.
I. Rojas-García and F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez
A Practical Implementation of Quantum LSTM Using Qiskit and PyTorch Conference
SOCO 20th Conference on Computing Systems and Applications, Communications in Computer and Information Science 2025.
Abstract | Links | BibTeX | Tags: quantum computing
@conference{SOCO25_Rojas,
title = {A Practical Implementation of Quantum LSTM Using Qiskit and PyTorch},
author = {I. Rojas-García and F. Rodríguez-Díaz and D. Gutiérrez-Avilés and A. Troncoso and F. Martínez-Álvarez},
url = {https://link.springer.com/chapter/10.1007/978-3-032-19763-4_32},
doi = {https://doi.org/10.1007/978-3-032-19763-4_32},
year = {2025},
date = {2025-06-15},
urldate = {2025-06-15},
booktitle = {SOCO 20th Conference on Computing Systems and Applications},
pages = {347–356},
series = {Communications in Computer and Information Science},
abstract = {Quantum computing is emerging as a promising tool to enhance classical machine learning models, especially in areas that involve complex temporal dynamics and high-dimensional data. This paper presents a practical implementation of a Quantum Long Short-Term Memory network. This hybrid architecture integrates parameterized quantum circuits into the gating mechanisms of a classical Long Short-Term Memory network. Built using PyTorch and Qiskit, our model enables end-to-end simulation, training, and analysis of quantum-enhanced recurrent networks. We design modular quantum circuits that interact with the Long Short-Term Memory’s internal states and evaluate their performance in processing synthetic sequential data. The architecture includes visualization and diagnostic tools to explore hidden state evolution, final memory values, and gate behaviors. Furthermore, we introduce an automated performance optimization framework that compares multiple Quantum Long Short-Term Memory network configurations using varying key quantum parameters such as qubit count and entanglement depth. Our findings demonstrate the feasibility of constructing interpretable, tunable quantum-classical sequence models using widely available tools. This work contributes a fully reproducible platform for experimentation and lays the groundwork for future research in scalable quantum recurrent architectures.},
keywords = {quantum computing},
pubstate = {published},
tppubtype = {conference}
}
Quantum computing is emerging as a promising tool to enhance classical machine learning models, especially in areas that involve complex temporal dynamics and high-dimensional data. This paper presents a practical implementation of a Quantum Long Short-Term Memory network. This hybrid architecture integrates parameterized quantum circuits into the gating mechanisms of a classical Long Short-Term Memory network. Built using PyTorch and Qiskit, our model enables end-to-end simulation, training, and analysis of quantum-enhanced recurrent networks. We design modular quantum circuits that interact with the Long Short-Term Memory’s internal states and evaluate their performance in processing synthetic sequential data. The architecture includes visualization and diagnostic tools to explore hidden state evolution, final memory values, and gate behaviors. Furthermore, we introduce an automated performance optimization framework that compares multiple Quantum Long Short-Term Memory network configurations using varying key quantum parameters such as qubit count and entanglement depth. Our findings demonstrate the feasibility of constructing interpretable, tunable quantum-classical sequence models using widely available tools. This work contributes a fully reproducible platform for experimentation and lays the groundwork for future research in scalable quantum recurrent architectures.
E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez
Communications in Computer and Information Science, Springer, vol. 2806, 2025, ISBN: 978-3-032-19763-4.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{SOCO2025_proceedings,
title = {Proceedings of the 20th International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2025) Salamanca, Spain, October 16-17, 2025},
author = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez},
editor = {E. Corchado and H. Quintián and A. Troncoso and H. Pérez-García and E. Jove-Pérez and P. García-Bringas and P. Fosci and F. Martínez-Álvarez},
url = {https://link.springer.com/book/10.1007/978-3-032-19763-4},
doi = {https://doi.org/10.1007/978-3-032-19763-4},
isbn = {978-3-032-19763-4},
year = {2025},
date = {2025-06-14},
volume = {2806},
publisher = {Communications in Computer and Information Science, Springer},
series = {Communications in Computer and Information Science},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
N. Ullah and F. Guzmán-Aroca and F. Martínez-Álvarez and I. De Falco and G. Sannino
A Novel Explainable AI Framework for Medical Image Classification Integrating Statistical, Visual, and Rule-Based Methods Journal Article
In: Medical Image Analysis, vol. 105, pp. 103665, 2025.
Abstract | Links | BibTeX | Tags: association rules, deep learning, feature selection, XAI
@article{ULLAH25,
title = {A Novel Explainable AI Framework for Medical Image Classification Integrating Statistical, Visual, and Rule-Based Methods},
author = {N. Ullah and F. Guzmán-Aroca and F. Martínez-Álvarez and I. De Falco and G. Sannino},
url = {https://www.sciencedirect.com/science/article/pii/S1361841525002129},
doi = {https://doi.org/10.1016/j.media.2025.103665},
year = {2025},
date = {2025-06-06},
urldate = {2025-06-06},
journal = {Medical Image Analysis},
volume = {105},
pages = {103665},
abstract = {Artificial intelligence and deep learning are powerful tools for extracting knowledge from large datasets, particularly in healthcare. However, their black-box nature raises interpretability concerns, especially in highstakes applications. Existing eXplainable Artificial Intelligence methods often focus solely on visualization or rule-based explanations, limiting interpretability’s depth and clarity. This work proposes a novel explainable AI method specifically designed for medical image analysis, integrating statistical, visual, and rule-based explanations to improve transparency in deep learning models. Statistical features are derived from deep features extracted using a custom Mobilenetv2 model. A two-step feature selection method—zero-based filtering with mutual importance selection—ranks and refines these features. Decision tree and RuleFit models
are employed to classify data and extract human-readable rules. Additionally, a novel statistical feature map overlay visualization generates heatmap-like representations of three key statistical measures (mean, skewness, and entropy), providing both localized and quantifiable visual explanations of model decisions. The proposed method has been validated on five medical imaging datasets—COVID-19 radiography, ultrasound
breast cancer, brain tumour magnetic resonance imaging, lung and colon cancer histopathological, and glaucoma images—with results confirmed by medical experts, demonstrating its effectiveness in enhancing interpretability for medical image classification tasks.},
keywords = {association rules, deep learning, feature selection, XAI},
pubstate = {published},
tppubtype = {article}
}
Artificial intelligence and deep learning are powerful tools for extracting knowledge from large datasets, particularly in healthcare. However, their black-box nature raises interpretability concerns, especially in highstakes applications. Existing eXplainable Artificial Intelligence methods often focus solely on visualization or rule-based explanations, limiting interpretability’s depth and clarity. This work proposes a novel explainable AI method specifically designed for medical image analysis, integrating statistical, visual, and rule-based explanations to improve transparency in deep learning models. Statistical features are derived from deep features extracted using a custom Mobilenetv2 model. A two-step feature selection method—zero-based filtering with mutual importance selection—ranks and refines these features. Decision tree and RuleFit models
are employed to classify data and extract human-readable rules. Additionally, a novel statistical feature map overlay visualization generates heatmap-like representations of three key statistical measures (mean, skewness, and entropy), providing both localized and quantifiable visual explanations of model decisions. The proposed method has been validated on five medical imaging datasets—COVID-19 radiography, ultrasound
breast cancer, brain tumour magnetic resonance imaging, lung and colon cancer histopathological, and glaucoma images—with results confirmed by medical experts, demonstrating its effectiveness in enhancing interpretability for medical image classification tasks.
are employed to classify data and extract human-readable rules. Additionally, a novel statistical feature map overlay visualization generates heatmap-like representations of three key statistical measures (mean, skewness, and entropy), providing both localized and quantifiable visual explanations of model decisions. The proposed method has been validated on five medical imaging datasets—COVID-19 radiography, ultrasound
breast cancer, brain tumour magnetic resonance imaging, lung and colon cancer histopathological, and glaucoma images—with results confirmed by medical experts, demonstrating its effectiveness in enhancing interpretability for medical image classification tasks.
A. Gil-Gamboa and J. F. Torres and F. Martínez-Álvarez and A. Troncoso
Energy-efficient transfer learning for water consumption forecasting Journal Article
In: Sustainable Computing: Informatics and Systems, vol. 46, pp. 101130, 2025.
Abstract | Links | BibTeX | Tags: deep learning, forecasting, time series, transfer learning
@article{GIL-GAMBOA25,
title = {Energy-efficient transfer learning for water consumption forecasting},
author = {A. Gil-Gamboa and J. F. Torres and F. Martínez-Álvarez and A. Troncoso},
url = {https://www.sciencedirect.com/science/article/pii/S2210537925000502},
doi = {https://doi.org/10.1016/j.suscom.2025.101130},
year = {2025},
date = {2025-05-07},
urldate = {2025-05-07},
journal = {Sustainable Computing: Informatics and Systems},
volume = {46},
pages = {101130},
abstract = {Artificial intelligence is expanding at an unprecedented rate due to the numerous advantages it provides to all types of businesses and industries. Water utilities are adopting artificial intelligence models to optimize water management in cities nowadays. However, the substantial computational demands of artificial intelligence present challenges, particularly regarding energy consumption and environmental impact. This paper addresses this problem by proposing a transfer learning approach for water consumption forecasting that reduces computational time, energy usage, and CO$_2$ emissions. The proposed methodology consists in developing a transfer learning approach based on a deep learning model already trained for a task with similar characteristics such as predicting electricity consumption. Thus, a pre-trained deep learning model designed for electricity consumption prediction is adapted to the water consumption domain, leveraging shared characteristics between these tasks. Experiments are conducted to determine the optimal amount of knowledge transfer and compare the performance of this approach with other state-of-the-art time-series forecasting models. Using real data from a water company in Spain, the transfer learning model achieves a similar or better accuracy than the other methods, while demonstrating significantly lower computational times, energy consumption and CO2 emissions. In addition, a scalability analysis has been conducted leading to the conclusion that the proposed transfer learning model is highly suitable to deal with big data. These findings highlight the potential of transfer learning as a sustainable and scalable solution for big data challenges in water management systems.},
keywords = {deep learning, forecasting, time series, transfer learning},
pubstate = {published},
tppubtype = {article}
}
Artificial intelligence is expanding at an unprecedented rate due to the numerous advantages it provides to all types of businesses and industries. Water utilities are adopting artificial intelligence models to optimize water management in cities nowadays. However, the substantial computational demands of artificial intelligence present challenges, particularly regarding energy consumption and environmental impact. This paper addresses this problem by proposing a transfer learning approach for water consumption forecasting that reduces computational time, energy usage, and CO$_2$ emissions. The proposed methodology consists in developing a transfer learning approach based on a deep learning model already trained for a task with similar characteristics such as predicting electricity consumption. Thus, a pre-trained deep learning model designed for electricity consumption prediction is adapted to the water consumption domain, leveraging shared characteristics between these tasks. Experiments are conducted to determine the optimal amount of knowledge transfer and compare the performance of this approach with other state-of-the-art time-series forecasting models. Using real data from a water company in Spain, the transfer learning model achieves a similar or better accuracy than the other methods, while demonstrating significantly lower computational times, energy consumption and CO2 emissions. In addition, a scalability analysis has been conducted leading to the conclusion that the proposed transfer learning model is highly suitable to deal with big data. These findings highlight the potential of transfer learning as a sustainable and scalable solution for big data challenges in water management systems.
D. Gutiérrez-Avilés and M. J. Jiménez-Navarro and J. F. Torres and F. Martínez-Álvarez
MetaGen: A framework for metaheuristic development and hyperparameter optimization in machine and deep learning Journal Article
In: Neurocomputing, vol. 637, pp. 130046, 2025.
Abstract | Links | BibTeX | Tags: big data, deep learning
@article{GUTIERREZ-AVILES25,
title = {MetaGen: A framework for metaheuristic development and hyperparameter optimization in machine and deep learning},
author = {D. Gutiérrez-Avilés and M. J. Jiménez-Navarro and J. F. Torres and F. Martínez-Álvarez},
url = {https://www.sciencedirect.com/science/article/pii/S0925231225007180},
doi = {https://doi.org/10.1016/j.neucom.2025.130046},
year = {2025},
date = {2025-03-15},
urldate = {2025-03-15},
journal = {Neurocomputing},
volume = {637},
pages = {130046},
abstract = {Hyperparameter optimization is a pivotal step in enhancing model performance within machine learning. Traditionally, this challenge is addressed through metaheuristics, which efficiently explore large search spaces to uncover optimal solutions. However, implementing these techniques can be complex without adequate development tools, which is the primary focus of this paper. Hence, we introduce MetaGen, a novel Python package designed to provide a comprehensive framework for developing and evaluating metaheuristic algorithms. MetaGen follows best practices in Python design, ensuring minimalistic code implementation, intuitive comprehension, and full flexibility in solution representation. The package defines two distinct user roles: Developers, responsible for algorithm implementation for hyperparameter optimization, and Solvers, who leverage pre-implemented metaheuristics to address optimization problems. Beyond algorithm implementation, MetaGen facilitates benchmarking through built-in test functions, ensuring standardized performance comparisons. It also provides automated reporting and visualization tools to analyze optimization progress and outcomes effectively. Furthermore, its modular design allows distribution
and integration into existing machine learning workflows. Several illustrative use cases are presented to demonstrate its adaptability and efficacy. The package, along with code, a user manual, and supplementary materials, is available at: https://github.com/Data-Science-Big-Data-Research-Lab/MetaGen.},
keywords = {big data, deep learning},
pubstate = {published},
tppubtype = {article}
}
Hyperparameter optimization is a pivotal step in enhancing model performance within machine learning. Traditionally, this challenge is addressed through metaheuristics, which efficiently explore large search spaces to uncover optimal solutions. However, implementing these techniques can be complex without adequate development tools, which is the primary focus of this paper. Hence, we introduce MetaGen, a novel Python package designed to provide a comprehensive framework for developing and evaluating metaheuristic algorithms. MetaGen follows best practices in Python design, ensuring minimalistic code implementation, intuitive comprehension, and full flexibility in solution representation. The package defines two distinct user roles: Developers, responsible for algorithm implementation for hyperparameter optimization, and Solvers, who leverage pre-implemented metaheuristics to address optimization problems. Beyond algorithm implementation, MetaGen facilitates benchmarking through built-in test functions, ensuring standardized performance comparisons. It also provides automated reporting and visualization tools to analyze optimization progress and outcomes effectively. Furthermore, its modular design allows distribution
and integration into existing machine learning workflows. Several illustrative use cases are presented to demonstrate its adaptability and efficacy. The package, along with code, a user manual, and supplementary materials, is available at: https://github.com/Data-Science-Big-Data-Research-Lab/MetaGen.
and integration into existing machine learning workflows. Several illustrative use cases are presented to demonstrate its adaptability and efficacy. The package, along with code, a user manual, and supplementary materials, is available at: https://github.com/Data-Science-Big-Data-Research-Lab/MetaGen.
E. T. Habtemariam and M. Martínez-Ballesteros and A. Troncoso and F. Martínez-Álvarez
A novel approach based on clustering and optimized ensemble deep learning for energy consumption forecasting in Ethiopia Journal Article
In: Neurocomputing, vol. 637, pp. 130027, 2025.
Abstract | Links | BibTeX | Tags: clustering, deep learning, energy, forecasting
@article{HABTEMARIAM25,
title = {A novel approach based on clustering and optimized ensemble deep learning for energy consumption forecasting in Ethiopia},
author = {E. T. Habtemariam and M. Martínez-Ballesteros and A. Troncoso and F. Martínez-Álvarez},
url = {https://www.sciencedirect.com/science/article/pii/S092523122500699X},
doi = {https://doi.org/10.1016/j.neucom.2025.130027},
year = {2025},
date = {2025-03-13},
urldate = {2025-03-13},
journal = {Neurocomputing},
volume = {637},
pages = {130027},
abstract = {Predicting energy consumption accurately is crucial for optimizing energy management strategies and achieving sustainability goals. Traditional methods often struggle with the complexity of energy consumption patterns, particularly in developing regions such as Ethiopia, where unique challenges exist. This study proposes an ensemble deep learning approach that integrates multiple models to enhance prediction accuracy.
Additionally, as a previous step, a clustering process has been applied to discover different groups of customers. Our method combines deep learning architectures, including Gated Recurrent Units, Long Short-Term Memory, and Convolutional Neural Networks, within an optimized ensemble with weights computed with the Coronavirus Optimization Algorithm. This approach aims to leverage the strengths of each model
to produce robust and reliable predictions. We demonstrate that our ensemble approach yields competitive results, outperforming individual models within the ensemble. By integrating diverse models, our framework captures nuanced patterns in energy consumption data more effectively, contributing to improved prediction accuracy. Furthermore, we validate the effectiveness of our approach using three distinct datasets from Ethiopia for three different customer clusters. These datasets represent different regions and consumption profiles within the country, ensuring the robustness and generalizability of our proposed methodology.},
keywords = {clustering, deep learning, energy, forecasting},
pubstate = {published},
tppubtype = {article}
}
Predicting energy consumption accurately is crucial for optimizing energy management strategies and achieving sustainability goals. Traditional methods often struggle with the complexity of energy consumption patterns, particularly in developing regions such as Ethiopia, where unique challenges exist. This study proposes an ensemble deep learning approach that integrates multiple models to enhance prediction accuracy.
Additionally, as a previous step, a clustering process has been applied to discover different groups of customers. Our method combines deep learning architectures, including Gated Recurrent Units, Long Short-Term Memory, and Convolutional Neural Networks, within an optimized ensemble with weights computed with the Coronavirus Optimization Algorithm. This approach aims to leverage the strengths of each model
to produce robust and reliable predictions. We demonstrate that our ensemble approach yields competitive results, outperforming individual models within the ensemble. By integrating diverse models, our framework captures nuanced patterns in energy consumption data more effectively, contributing to improved prediction accuracy. Furthermore, we validate the effectiveness of our approach using three distinct datasets from Ethiopia for three different customer clusters. These datasets represent different regions and consumption profiles within the country, ensuring the robustness and generalizability of our proposed methodology.
Additionally, as a previous step, a clustering process has been applied to discover different groups of customers. Our method combines deep learning architectures, including Gated Recurrent Units, Long Short-Term Memory, and Convolutional Neural Networks, within an optimized ensemble with weights computed with the Coronavirus Optimization Algorithm. This approach aims to leverage the strengths of each model
to produce robust and reliable predictions. We demonstrate that our ensemble approach yields competitive results, outperforming individual models within the ensemble. By integrating diverse models, our framework captures nuanced patterns in energy consumption data more effectively, contributing to improved prediction accuracy. Furthermore, we validate the effectiveness of our approach using three distinct datasets from Ethiopia for three different customer clusters. These datasets represent different regions and consumption profiles within the country, ensuring the robustness and generalizability of our proposed methodology.
A. R. Troncoso-García and M. Martínez-Ballesteros and F. Martínez-Álvarez and A. Troncoso
A new metric based on association rules to assess explainability techniques for time series forecasting Journal Article
In: IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 5, pp. 4140-4155, 2025.
Abstract | Links | BibTeX | Tags: association rules, forecasting, time series, XAI
@article{TRONCOSO-GARCIA25,
title = {A new metric based on association rules to assess explainability techniques for time series forecasting},
author = {A. R. Troncoso-García and M. Martínez-Ballesteros and F. Martínez-Álvarez and A. Troncoso},
url = {https://ieeexplore.ieee.org/document/10879535},
doi = {10.1109/TPAMI.2025.3540513},
year = {2025},
date = {2025-02-11},
urldate = {2025-02-11},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
volume = {47},
number = {5},
pages = {4140-4155},
abstract = {This paper introduces a new, model-independent, metric, called RExQUAL, for quantifying the quality of explanations provided by attribution-based explainable artificial intelligence techniques and compare them. The underlying idea is based on feature attribution, using a subset of the ranking of the attributes highlighted by a model-agnostic explainable method in a forecasting task. Then, association rules are generated using these key attributes as input data. Novel metrics, including global support and confidence, are proposed to assess the joint quality of generated rules. Finally, the quality of the explanations is calculated based on a wise and comprehensive combination of the association rules global metrics. The proposed method integrates local explanations through attribution-based approaches for evaluation and feature selection with global explanations for the entire dataset. This paper rigorously evaluates the new metric by comparing three explainability techniques: the widely used SHAP and LIME, and the novel methodology RULEx. The experimental design includes predicting time series of different natures, including univariate and multivariate, through deep learning models. The results underscore the efficacy and versatility of the proposed methodology as a quantitative framework for evaluating and comparing explainable techniques.},
keywords = {association rules, forecasting, time series, XAI},
pubstate = {published},
tppubtype = {article}
}
This paper introduces a new, model-independent, metric, called RExQUAL, for quantifying the quality of explanations provided by attribution-based explainable artificial intelligence techniques and compare them. The underlying idea is based on feature attribution, using a subset of the ranking of the attributes highlighted by a model-agnostic explainable method in a forecasting task. Then, association rules are generated using these key attributes as input data. Novel metrics, including global support and confidence, are proposed to assess the joint quality of generated rules. Finally, the quality of the explanations is calculated based on a wise and comprehensive combination of the association rules global metrics. The proposed method integrates local explanations through attribution-based approaches for evaluation and feature selection with global explanations for the entire dataset. This paper rigorously evaluates the new metric by comparing three explainability techniques: the widely used SHAP and LIME, and the novel methodology RULEx. The experimental design includes predicting time series of different natures, including univariate and multivariate, through deep learning models. The results underscore the efficacy and versatility of the proposed methodology as a quantitative framework for evaluating and comparing explainable techniques.
R. Scitovski and K. Sabo and D. Grahovac and F. Martínez-Álvarez and S. Ungar
A partitioning incremental algorithm using adaptive Mahalanobis fuzzy clustering and identifying the most appropriate partition Journal Article
In: Pattern Analysis and Applications, vol. 28, pp. 3, 2025.
Abstract | Links | BibTeX | Tags: clustering
@article{SCITOVSKI25,
title = {A partitioning incremental algorithm using adaptive Mahalanobis fuzzy clustering and identifying the most appropriate partition},
author = {R. Scitovski and K. Sabo and D. Grahovac and F. Martínez-Álvarez and S. Ungar},
url = {https://link.springer.com/article/10.1007/s10044-024-01360-2},
doi = {https://doi.org/10.1007/s10044-024-01360-2},
year = {2025},
date = {2025-01-02},
journal = {Pattern Analysis and Applications},
volume = {28},
pages = {3},
abstract = {This paper deals with the problem of determining the most appropriate number of clusters in a fuzzy Mahalanobis partition.
First, a new fuzzy Mahalanobis incremental algorithm is constructed to search for an optimal fuzzy Mahalanobis
partition with 2, 3, ... clusters. Among these partitions, selecting the one with the most appropriate number of clusters
is based on appropriately modified existing fuzzy indexes. In addition, the Fuzzy Mahalanobis Minimal Distance index
is defined as a natural extension of the recently proposed Mahalanobis Minimal Distance index for non-fuzzy clustering.
The new fuzzy Mahalanobis incremental algorithm was tested on several artificial data sets and the color image segmentation
problems from real-world applications: art images, nature photography images, and medical images. The algorithm
includes multiple usage of the global optimization algorithm DIRECT. But unlike previously known fuzzy Mahalanobis
indexes, the proposed Fuzzy Mahalanobis Minimal Distance index ensures accurate results even when applied to complex
real-world applications. A possible disadvantage could be the need for longer CPU time. Furthermore, besides effective
identification of the partition with the most appropriate number of clusters, it is shown how to use the proposed Fuzzy
Mahalanobis Minimal Distance index to search for an acceptable partition, which proved particularly useful in the abovementioned
real-world applications.},
keywords = {clustering},
pubstate = {published},
tppubtype = {article}
}
This paper deals with the problem of determining the most appropriate number of clusters in a fuzzy Mahalanobis partition.
First, a new fuzzy Mahalanobis incremental algorithm is constructed to search for an optimal fuzzy Mahalanobis
partition with 2, 3, ... clusters. Among these partitions, selecting the one with the most appropriate number of clusters
is based on appropriately modified existing fuzzy indexes. In addition, the Fuzzy Mahalanobis Minimal Distance index
is defined as a natural extension of the recently proposed Mahalanobis Minimal Distance index for non-fuzzy clustering.
The new fuzzy Mahalanobis incremental algorithm was tested on several artificial data sets and the color image segmentation
problems from real-world applications: art images, nature photography images, and medical images. The algorithm
includes multiple usage of the global optimization algorithm DIRECT. But unlike previously known fuzzy Mahalanobis
indexes, the proposed Fuzzy Mahalanobis Minimal Distance index ensures accurate results even when applied to complex
real-world applications. A possible disadvantage could be the need for longer CPU time. Furthermore, besides effective
identification of the partition with the most appropriate number of clusters, it is shown how to use the proposed Fuzzy
Mahalanobis Minimal Distance index to search for an acceptable partition, which proved particularly useful in the abovementioned
real-world applications.
First, a new fuzzy Mahalanobis incremental algorithm is constructed to search for an optimal fuzzy Mahalanobis
partition with 2, 3, ... clusters. Among these partitions, selecting the one with the most appropriate number of clusters
is based on appropriately modified existing fuzzy indexes. In addition, the Fuzzy Mahalanobis Minimal Distance index
is defined as a natural extension of the recently proposed Mahalanobis Minimal Distance index for non-fuzzy clustering.
The new fuzzy Mahalanobis incremental algorithm was tested on several artificial data sets and the color image segmentation
problems from real-world applications: art images, nature photography images, and medical images. The algorithm
includes multiple usage of the global optimization algorithm DIRECT. But unlike previously known fuzzy Mahalanobis
indexes, the proposed Fuzzy Mahalanobis Minimal Distance index ensures accurate results even when applied to complex
real-world applications. A possible disadvantage could be the need for longer CPU time. Furthermore, besides effective
identification of the partition with the most appropriate number of clusters, it is shown how to use the proposed Fuzzy
Mahalanobis Minimal Distance index to search for an acceptable partition, which proved particularly useful in the abovementioned
real-world applications.
P. Casas-Gómez and J. F. Torres and J. C. Linares and A. Troncoso and F. Martínez-Álvarez
Forecasting basal area increment in forest ecosystems using deep learning: A multi-species analysis in the Himalayas Journal Article
In: Ecological Informatics, vol. 85, pp. 102951, 2025.
Abstract | Links | BibTeX | Tags: deep learning, time series
@article{CASAS-GOMEZ25,
title = {Forecasting basal area increment in forest ecosystems using deep learning: A multi-species analysis in the Himalayas},
author = {P. Casas-Gómez and J. F. Torres and J. C. Linares and A. Troncoso and F. Martínez-Álvarez},
url = {https://www.sciencedirect.com/science/article/pii/S157495412400493X?via%3Dihub},
doi = {https://doi.org/10.1016/j.ecoinf.2024.102951},
year = {2025},
date = {2025-01-01},
journal = {Ecological Informatics},
volume = {85},
pages = {102951},
abstract = {This study addresses the task of forecasting Basal Area Increment trends in forest ecosystems, which is essential for conservation and biodiversity management, particularly in the context of climate change. Traditional forecasting techniques, such as Linear Mixed Models, Random Forest and standard Artificial Neural Networks, often fail to account for the time-dependent nature of tree growth and utilize simple architectures. To overcome these limitations, we introduce the use of two different Deep Learning models: the Long Short-Term Memory network and the Temporal Convolutional Neural Network, which capture the temporal dependencies of growth by incorporating lagged Basal Area Increment values. Our methodology includes rigorous hyperparameter tuning to optimize the Deep Learning models’ architecture. We evaluate the models’ performance across 15 species in the Himalayan region, individually and collectively, using temperature and precipitation data as predictors. The Deep Learning model significantly outperforms state-of-the-art techniques, achieving the lowest Root Mean Squared Error (7.407 for LSTM and 6.202 for TCNN), highest 𝑅2 (0.495 for LSTM an0.585 for TCNN) and lowest Mean Absolute Percentage Error values (37.653 for LSTM and 34.296 for TCNN). These findings highlight the potential of Deep Learning networks to provide accurate and reliable Basal AreaIncrement forecasts, offering valuable insights for forest management and conservation efforts in the face of ongoing climate change.},
keywords = {deep learning, time series},
pubstate = {published},
tppubtype = {article}
}
This study addresses the task of forecasting Basal Area Increment trends in forest ecosystems, which is essential for conservation and biodiversity management, particularly in the context of climate change. Traditional forecasting techniques, such as Linear Mixed Models, Random Forest and standard Artificial Neural Networks, often fail to account for the time-dependent nature of tree growth and utilize simple architectures. To overcome these limitations, we introduce the use of two different Deep Learning models: the Long Short-Term Memory network and the Temporal Convolutional Neural Network, which capture the temporal dependencies of growth by incorporating lagged Basal Area Increment values. Our methodology includes rigorous hyperparameter tuning to optimize the Deep Learning models’ architecture. We evaluate the models’ performance across 15 species in the Himalayan region, individually and collectively, using temperature and precipitation data as predictors. The Deep Learning model significantly outperforms state-of-the-art techniques, achieving the lowest Root Mean Squared Error (7.407 for LSTM and 6.202 for TCNN), highest 𝑅2 (0.495 for LSTM an0.585 for TCNN) and lowest Mean Absolute Percentage Error values (37.653 for LSTM and 34.296 for TCNN). These findings highlight the potential of Deep Learning networks to provide accurate and reliable Basal AreaIncrement forecasts, offering valuable insights for forest management and conservation efforts in the face of ongoing climate change.
L. Melgar-García and D. Gutiérrez-Avilés and C. Rubio-Escudero and A. Troncoso
Online forecasting using neighbor-based incremental learning for electricity markets Journal Article
In: Neural Computing and Applications, vol. 37, pp. 22923–22940, 2025.
Links | BibTeX | Tags: energy, IoT, time series
@article{Melgar2025,
title = {Online forecasting using neighbor-based incremental learning for electricity markets},
author = {L. Melgar-García and D. Gutiérrez-Avilés and C. Rubio-Escudero and A. Troncoso },
url = {https://link.springer.com/article/10.1007/s00521-024-10876-x},
doi = {https://doi.org/10.1007/s00521-024-10876-x},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Neural Computing and Applications},
volume = {37},
pages = {22923–22940},
keywords = {energy, IoT, time series},
pubstate = {published},
tppubtype = {article}
}
A. M. Chacón-Maldonado and L. Melgar-García and G. Asencio-Cortés and A. Troncoso
A novel method based on hybrid deep learning with explainability for olive fruit pest forecasting Journal Article
In: Neural Computing and Applications, vol. 37, pp. 3245-3264, 2025.
Links | BibTeX | Tags: deep learning, precision agriculture, XAI
@article{Chacon2025,
title = {A novel method based on hybrid deep learning with explainability for olive fruit pest forecasting},
author = {A. M. Chacón-Maldonado and L. Melgar-García and G. Asencio-Cortés and A. Troncoso},
url = {https://link.springer.com/article/10.1007/s00521-024-10731-z},
doi = {https://doi.org/10.1007/s00521-024-10731-z},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Neural Computing and Applications},
volume = {37},
pages = {3245-3264},
keywords = {deep learning, precision agriculture, XAI},
pubstate = {published},
tppubtype = {article}
}
A. López-Fernández and F. Divina and F. A. Gómez-Vela and M. García-Torres
Data mining for enhancing learning and assessment to a microcompetence-based methodology in higher education Journal Article
In: IEEE Revista Iberoamericana de Tecnologias del Aprendizaje, 2025.
Abstract | Links | BibTeX | Tags: clustering, education
@article{lopez2025data,
title = {Data mining for enhancing learning and assessment to a microcompetence-based methodology in higher education},
author = { A. López-Fernández and F. Divina and F. A. Gómez-Vela and M. García-Torres},
doi = {10.1109/RITA.2025.3532879},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {IEEE Revista Iberoamericana de Tecnologias del Aprendizaje},
publisher = {IEEE},
abstract = {This work introduces an innovative teaching methodology based on microcompetences applied in a higher education context. The intervention involved creating a repository of practical case studies in the form of quizzes and integrating microcompetences
into each course activity. The digital tool Sapiens was used to identify learning deficiencies and provide both collective and individualized feedback. The results indicate a significant increase in student participation and academic performance compared to
previous years. Furthermore, students voluntarily used virtual teaching modalities to reinforce their knowledge, particularly in more complex areas. Data mining techniques identified performance patterns among students, highlighting the methodology’s
effectiveness in improving both transversal and specific competences. The study’s findings underscore the importance of implementing microcompetency-based methodologies in higher education to enhance the quality of learning and continuous assessment. This
approach not only facilitated a deeper understanding of course content but also promoted critical thinking, abstract reasoning, and interpersonal skills, preparing students for future academic and professional challenges. Additionally, the flexibility and
adaptability of the digital tools used provided a seamless transition across different teaching modalities, such as in-person, hybrid, and online formats. Thus, the implementation of this innovative methodology has demonstrated its potential to significantly
improve student engagement, participation, and academic success, thereby contributing to a more effective and comprehensive educational experience in higher education. url = https://ieeexplore.ieee.org/abstract/document/10849581},
keywords = {clustering, education},
pubstate = {published},
tppubtype = {article}
}
This work introduces an innovative teaching methodology based on microcompetences applied in a higher education context. The intervention involved creating a repository of practical case studies in the form of quizzes and integrating microcompetences
into each course activity. The digital tool Sapiens was used to identify learning deficiencies and provide both collective and individualized feedback. The results indicate a significant increase in student participation and academic performance compared to
previous years. Furthermore, students voluntarily used virtual teaching modalities to reinforce their knowledge, particularly in more complex areas. Data mining techniques identified performance patterns among students, highlighting the methodology’s
effectiveness in improving both transversal and specific competences. The study’s findings underscore the importance of implementing microcompetency-based methodologies in higher education to enhance the quality of learning and continuous assessment. This
approach not only facilitated a deeper understanding of course content but also promoted critical thinking, abstract reasoning, and interpersonal skills, preparing students for future academic and professional challenges. Additionally, the flexibility and
adaptability of the digital tools used provided a seamless transition across different teaching modalities, such as in-person, hybrid, and online formats. Thus, the implementation of this innovative methodology has demonstrated its potential to significantly
improve student engagement, participation, and academic success, thereby contributing to a more effective and comprehensive educational experience in higher education. url = https://ieeexplore.ieee.org/abstract/document/10849581
into each course activity. The digital tool Sapiens was used to identify learning deficiencies and provide both collective and individualized feedback. The results indicate a significant increase in student participation and academic performance compared to
previous years. Furthermore, students voluntarily used virtual teaching modalities to reinforce their knowledge, particularly in more complex areas. Data mining techniques identified performance patterns among students, highlighting the methodology’s
effectiveness in improving both transversal and specific competences. The study’s findings underscore the importance of implementing microcompetency-based methodologies in higher education to enhance the quality of learning and continuous assessment. This
approach not only facilitated a deeper understanding of course content but also promoted critical thinking, abstract reasoning, and interpersonal skills, preparing students for future academic and professional challenges. Additionally, the flexibility and
adaptability of the digital tools used provided a seamless transition across different teaching modalities, such as in-person, hybrid, and online formats. Thus, the implementation of this innovative methodology has demonstrated its potential to significantly
improve student engagement, participation, and academic success, thereby contributing to a more effective and comprehensive educational experience in higher education. url = https://ieeexplore.ieee.org/abstract/document/10849581
M. García-Torres
Feature selection for high-dimensional data using a multivariate search space reduction strategy based scatter search Journal Article
In: Journal of Heuristics, vol. 31, no. 1, pp. 10, 2025.
Abstract | Links | BibTeX | Tags: feature selection
@article{garcia2025feature,
title = {Feature selection for high-dimensional data using a multivariate search space reduction strategy based scatter search},
author = {M. García-Torres},
doi = {10.1007/s10732-025-09550-9},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Journal of Heuristics},
volume = {31},
number = {1},
pages = {10},
publisher = {Springer},
abstract = {In feature selection, the increasing of the dimensionality and the complexity of feature interactions make the problem challenging. Furthermore, searching for an optimal subset of features from a high-dimensional feature space is known to be an
NP-hard problem. To improve the efficiency and effectiveness of the search algorithm, feature grouping has emerged as a way to reduce the search space by clustering features according to a measure. In this work we propose to reduce the search space by
applying a greedy algorithm, called Multivariate Greedy Predominant Groups Generator (MGPGG). MGPGG extends the idea of the Greedy Predominant Groups Generator (GPGG) algorithm by taking into account feature interaction among three or more features. For
this purpose, MGPGG uses the Multivariate Symmetrical Uncertainty (MSU) to group features that share information about the class label. We also propose a Scatter Search strategy that integrates MGPGG to find small subsets of features with high predictive power.
The proposed algorithm, called Multivariate Predominant Group-based Scatter Search (MPGSS), is tested on high-dimensional data from biomedical and text-mining fields. The proposal is compared with state-of-the-art feature selection strategies. Results show that
MPGSS is competitive since it is capable of finding small subsets of features while keeping high predictive classification models. url = https://link.springer.com/article/10.1007/s10732-025-09550-9},
keywords = {feature selection},
pubstate = {published},
tppubtype = {article}
}
In feature selection, the increasing of the dimensionality and the complexity of feature interactions make the problem challenging. Furthermore, searching for an optimal subset of features from a high-dimensional feature space is known to be an
NP-hard problem. To improve the efficiency and effectiveness of the search algorithm, feature grouping has emerged as a way to reduce the search space by clustering features according to a measure. In this work we propose to reduce the search space by
applying a greedy algorithm, called Multivariate Greedy Predominant Groups Generator (MGPGG). MGPGG extends the idea of the Greedy Predominant Groups Generator (GPGG) algorithm by taking into account feature interaction among three or more features. For
this purpose, MGPGG uses the Multivariate Symmetrical Uncertainty (MSU) to group features that share information about the class label. We also propose a Scatter Search strategy that integrates MGPGG to find small subsets of features with high predictive power.
The proposed algorithm, called Multivariate Predominant Group-based Scatter Search (MPGSS), is tested on high-dimensional data from biomedical and text-mining fields. The proposal is compared with state-of-the-art feature selection strategies. Results show that
MPGSS is competitive since it is capable of finding small subsets of features while keeping high predictive classification models. url = https://link.springer.com/article/10.1007/s10732-025-09550-9
NP-hard problem. To improve the efficiency and effectiveness of the search algorithm, feature grouping has emerged as a way to reduce the search space by clustering features according to a measure. In this work we propose to reduce the search space by
applying a greedy algorithm, called Multivariate Greedy Predominant Groups Generator (MGPGG). MGPGG extends the idea of the Greedy Predominant Groups Generator (GPGG) algorithm by taking into account feature interaction among three or more features. For
this purpose, MGPGG uses the Multivariate Symmetrical Uncertainty (MSU) to group features that share information about the class label. We also propose a Scatter Search strategy that integrates MGPGG to find small subsets of features with high predictive power.
The proposed algorithm, called Multivariate Predominant Group-based Scatter Search (MPGSS), is tested on high-dimensional data from biomedical and text-mining fields. The proposal is compared with state-of-the-art feature selection strategies. Results show that
MPGSS is competitive since it is capable of finding small subsets of features while keeping high predictive classification models. url = https://link.springer.com/article/10.1007/s10732-025-09550-9
M. García-Torres and F. Saucedo and F. Divina and S. Gómez
RFMSU: A multivariate symmetrical uncertainty based random forest Journal Article
In: Pattern Recognition, vol. 169, pp. 111939, 2025.
Abstract | Links | BibTeX | Tags: feature selection, pattern recognition
@article{garcia2025rfmsu,
title = {RFMSU: A multivariate symmetrical uncertainty based random forest},
author = {M. García-Torres and F. Saucedo and F. Divina and S. Gómez},
url = {https://www.sciencedirect.com/science/article/pii/S0031320325005990?via%3Dihub},
doi = {https://doi.org/10.1016/j.patcog.2025.111939},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Pattern Recognition},
volume = {169},
pages = {111939},
publisher = {Elsevier},
abstract = {Decision Trees (DTs) have become very popular classifiers due to their good performance and, most of all, their interpretability. In addition, the machine learning community is also paying attention to Random Forests (RFs) since they defy the interpretability-accuracy tradeoff. Most RFs strategies are based on univariate measures, a fact that may limit the capability of identifying the interaction among more than two features. In order to overcome this problem many multivariate approaches have been proposed. However, most of them are based on finding linear or non-linear combinations of features. In this work, we propose a novel univariate RF strategy that builds DTs using the Multivariate Symmetrical Uncertainty (MSU) measure as splitting criterion. The proposal, referred to as RF$_MSU$, was tested on high-dimensional datasets and compared to state-of-the-art univariate
and multivariate DTs and RFs classifiers. Results suggest that RF$_MSU$ is capable of finding simpler rules than other RFs approaches while keeping a high predictive power equivalent to that of multivariate approaches. The DT strategies considered obtained simpler models than RF$_MSU$, but at the expense of degrading the classifier. Thus, we can conclude that RFMS U is a RF-based classifier that achieves a good trade-off between the performance and the complexity of the model.},
keywords = {feature selection, pattern recognition},
pubstate = {published},
tppubtype = {article}
}
Decision Trees (DTs) have become very popular classifiers due to their good performance and, most of all, their interpretability. In addition, the machine learning community is also paying attention to Random Forests (RFs) since they defy the interpretability-accuracy tradeoff. Most RFs strategies are based on univariate measures, a fact that may limit the capability of identifying the interaction among more than two features. In order to overcome this problem many multivariate approaches have been proposed. However, most of them are based on finding linear or non-linear combinations of features. In this work, we propose a novel univariate RF strategy that builds DTs using the Multivariate Symmetrical Uncertainty (MSU) measure as splitting criterion. The proposal, referred to as RF$_MSU$, was tested on high-dimensional datasets and compared to state-of-the-art univariate
and multivariate DTs and RFs classifiers. Results suggest that RF$_MSU$ is capable of finding simpler rules than other RFs approaches while keeping a high predictive power equivalent to that of multivariate approaches. The DT strategies considered obtained simpler models than RF$_MSU$, but at the expense of degrading the classifier. Thus, we can conclude that RFMS U is a RF-based classifier that achieves a good trade-off between the performance and the complexity of the model.
and multivariate DTs and RFs classifiers. Results suggest that RF$_MSU$ is capable of finding simpler rules than other RFs approaches while keeping a high predictive power equivalent to that of multivariate approaches. The DT strategies considered obtained simpler models than RF$_MSU$, but at the expense of degrading the classifier. Thus, we can conclude that RFMS U is a RF-based classifier that achieves a good trade-off between the performance and the complexity of the model.
J. L. Vázquez Noguera and A. Torres-Hurtado and H. Gómez-Adorno and J. C. Mello-Román and E. J. Fleitas-Alvarez and F. F. Espinola Schulze and M. García-Torres and C. D. Méndez Gaona and P. E. Gardel Sotomayor and S. Vázquez Noguera and N. E. Zaracho Amarilla and O. W. Gamarra Esquivel
Mammography Reporting Dataset with BI-RADS System for Natural Language Processing Applications: Addressing Public Data Gaps in Spanish Journal Article
In: Data in Brief, vol. 61, pp. 111761, 2025.
Abstract | Links | BibTeX | Tags: large language model
@article{vazquez2025mammography,
title = {Mammography Reporting Dataset with BI-RADS System for Natural Language Processing Applications: Addressing Public Data Gaps in Spanish},
author = { J. L. Vázquez Noguera and A. Torres-Hurtado and H. Gómez-Adorno and J. C. Mello-Román and E. J. Fleitas-Alvarez and F. F. Espinola Schulze and M. García-Torres and C. D. Méndez Gaona and P. E. Gardel Sotomayor and S. Vázquez Noguera and N. E. Zaracho Amarilla and O. W. Gamarra Esquivel},
url = {https://www.sciencedirect.com/science/article/pii/S2352340925004883?via%3Dihub},
doi = {https://doi.org/10.1016/j.dib.2025.111761},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Data in Brief},
volume = {61},
pages = {111761},
publisher = {Elsevier},
abstract = {Applying Natural Language Processing (NLP) to clinical reports is important for automating the analysis and classification of clinical data, improving diagnostic accuracy, and enhancing healthcare workflows. This article presents a dataset derived from mammography reports written in Spanish collected across multiple medical units operated by the Oxades company in Paraguay. The dataset contains 4,357 records and 15 variables, including the text of the complete report and also each of its sections separately (clinical observations, diagnostic conclusions, follow-up recommendations), and the BI-RADS (Breast Imaging Reporting and Data System) classification assigned to each one of the reports. Additionally, the dataset includes metadata such as report IDs, dates, and patient information such as age, patient reasons for the analysis, last menstruation period, type of hormonal therapy received, family history and number of children. To ensure patient confidentiality, all identifiable data was removed, and the dataset was structured using automated segmentation and manual verification to ensure quality and transparency. This dataset is an invaluable resource for both medical and AI research communities. It provides real-world data for developing and testing NLP algorithms and machine learning models, specifically for automating BI-RADS classification and analyzing mammography reports.},
keywords = {large language model},
pubstate = {published},
tppubtype = {article}
}
Applying Natural Language Processing (NLP) to clinical reports is important for automating the analysis and classification of clinical data, improving diagnostic accuracy, and enhancing healthcare workflows. This article presents a dataset derived from mammography reports written in Spanish collected across multiple medical units operated by the Oxades company in Paraguay. The dataset contains 4,357 records and 15 variables, including the text of the complete report and also each of its sections separately (clinical observations, diagnostic conclusions, follow-up recommendations), and the BI-RADS (Breast Imaging Reporting and Data System) classification assigned to each one of the reports. Additionally, the dataset includes metadata such as report IDs, dates, and patient information such as age, patient reasons for the analysis, last menstruation period, type of hormonal therapy received, family history and number of children. To ensure patient confidentiality, all identifiable data was removed, and the dataset was structured using automated segmentation and manual verification to ensure quality and transparency. This dataset is an invaluable resource for both medical and AI research communities. It provides real-world data for developing and testing NLP algorithms and machine learning models, specifically for automating BI-RADS classification and analyzing mammography reports.
D. Rodríguez-Baena and F. Gómez-Vela and A. Lopez-Fernandez and M. García-Torres and F. Divina
BinRec: addressing data sparsity and cold-start challenges in recommender systems with biclustering Journal Article
In: Applied Intelligence, vol. 55, no. 12, pp. 830, 2025.
Abstract | Links | BibTeX | Tags: clustering, pattern recognition
@article{rodriguez2025binrec,
title = {BinRec: addressing data sparsity and cold-start challenges in recommender systems with biclustering},
author = {D. Rodríguez-Baena and F. Gómez-Vela and A. Lopez-Fernandez and M. García-Torres and F. Divina},
url = {https://link.springer.com/article/10.1007/s10489-025-06725-6},
doi = {10.1007/s10489-025-06725-6},
year = {2025},
date = {2025-01-01},
journal = {Applied Intelligence},
volume = {55},
number = {12},
pages = {830},
abstract = {Recommender Systems help users in making decision in different fields such as purchases or what movies to watch. User-Based Collaborative Filtering (UBCF) approach is one of the most commonly used techniques for developing these software tools. It is based on the idea that users who have previously shared similar tastes will almost certainly share similar tastes in the future. As a result, determining the nearest users to the one for whom recommendations are sought (active user) is critical. However, the massive growth of online commercial data has made this task especially difficult. As a result, Biclustering techniques have been used in recent years to perform a local search for the nearest users in subgroups of users with similar rating behaviour under a subgroup of items (biclusters), rather than searching the entire rating database. Nevertheless, due to the large size of these databases, the number of biclusters generated can be extremely high, making their processing very complex. In this paper we propose BinRec, a novel UBCF approach based on Biclustering. BinRec simplifies the search for neighbouring users by determining which ones are nearest to the active user based on the number of biclusters shared by the users. Experimental results show that BinRec outperforms other state-of-the-art recommender systems, with a remarkable improvement in environments with high data sparsity. The flexibility and scalability of the method position it as an efficient alternative for common collaborative filtering problems such as sparsity or cold-start.},
keywords = {clustering, pattern recognition},
pubstate = {published},
tppubtype = {article}
}
Recommender Systems help users in making decision in different fields such as purchases or what movies to watch. User-Based Collaborative Filtering (UBCF) approach is one of the most commonly used techniques for developing these software tools. It is based on the idea that users who have previously shared similar tastes will almost certainly share similar tastes in the future. As a result, determining the nearest users to the one for whom recommendations are sought (active user) is critical. However, the massive growth of online commercial data has made this task especially difficult. As a result, Biclustering techniques have been used in recent years to perform a local search for the nearest users in subgroups of users with similar rating behaviour under a subgroup of items (biclusters), rather than searching the entire rating database. Nevertheless, due to the large size of these databases, the number of biclusters generated can be extremely high, making their processing very complex. In this paper we propose BinRec, a novel UBCF approach based on Biclustering. BinRec simplifies the search for neighbouring users by determining which ones are nearest to the active user based on the number of biclusters shared by the users. Experimental results show that BinRec outperforms other state-of-the-art recommender systems, with a remarkable improvement in environments with high data sparsity. The flexibility and scalability of the method position it as an efficient alternative for common collaborative filtering problems such as sparsity or cold-start.
T. Vanhaeren and L. Cataneo and F. Divina and P. M. Martínez-García
Enhancing R-loop prediction with high-throughput sequencing data Journal Article
In: NAR Genomics and Bioinformatics, vol. 7, no. 2, pp. lqaf077, 2025.
Abstract | Links | BibTeX | Tags: big data, bioinformatics, pattern recognition
@article{vanhaeren2025enhancing,
title = {Enhancing R-loop prediction with high-throughput sequencing data},
author = {T. Vanhaeren and L. Cataneo and F. Divina and P. M. Martínez-García},
url = {https://academic.oup.com/nargab/article/7/2/lqaf077/8160316},
doi = {10.1093/nargab/lqaf077},
year = {2025},
date = {2025-01-01},
journal = {NAR Genomics and Bioinformatics},
volume = {7},
number = {2},
pages = {lqaf077},
abstract = {R-loops are three-stranded RNA and DNA hybrid structures that often occur in the genome and play important roles in a variety of cellular processes from bacteria to mammals. Sequencing methods profiling R-loops genome-wide have revealed that they can form co-transcriptionally at cell type specific genes and associate with specific chromatin states during cell differentiation and reprogramming. However, current computational methods for the prediction of R-loops rely solely on their DNA sequence properties, which precludes detection across cell types, tissues or developmental stages. Here, we conduct a machine learning approach that allows the prediction of mammalian cell type-specific R-loops using sequence information and high-throughput sequencing signals. Our predictive models are induced from human samples and achieve highly accurate predictions, with transcriptomics, DNA features, chromatin accessibility and the active gene body H3K36me3 epigenomic mark being the most informative datasets. We generate de novo virtual R-loop maps that show high concordance with experimental ones and capture cell type specificity. Our approach compares favorably to sequence-based methods and can be generalized to mouse datasets. Based on this, we generate virtual R-loop maps in 51 mammalian systems that are freely accessible to the scientific community.},
keywords = {big data, bioinformatics, pattern recognition},
pubstate = {published},
tppubtype = {article}
}
R-loops are three-stranded RNA and DNA hybrid structures that often occur in the genome and play important roles in a variety of cellular processes from bacteria to mammals. Sequencing methods profiling R-loops genome-wide have revealed that they can form co-transcriptionally at cell type specific genes and associate with specific chromatin states during cell differentiation and reprogramming. However, current computational methods for the prediction of R-loops rely solely on their DNA sequence properties, which precludes detection across cell types, tissues or developmental stages. Here, we conduct a machine learning approach that allows the prediction of mammalian cell type-specific R-loops using sequence information and high-throughput sequencing signals. Our predictive models are induced from human samples and achieve highly accurate predictions, with transcriptomics, DNA features, chromatin accessibility and the active gene body H3K36me3 epigenomic mark being the most informative datasets. We generate de novo virtual R-loop maps that show high concordance with experimental ones and capture cell type specificity. Our approach compares favorably to sequence-based methods and can be generalized to mouse datasets. Based on this, we generate virtual R-loop maps in 51 mammalian systems that are freely accessible to the scientific community.
T. Vanhaeren and A. R. Troncoso-García and J. F. Torres and F. Divina and P. M. Martínez-García
Application of XAI to the prediction of CTCF binding sites Journal Article
In: Results in Engineering, vol. 25, pp. 103776, 2025.
Abstract | Links | BibTeX | Tags: big data, bioinformatics, deep learning
@article{vanhaeren2025application,
title = {Application of XAI to the prediction of CTCF binding sites},
author = {T. Vanhaeren and A. R. Troncoso-García and J. F. Torres and F. Divina and P. M. Martínez-García},
url = {https://www.sciencedirect.com/science/article/pii/S259012302402019X},
doi = {10.1016/j.rineng.2024.103776},
year = {2025},
date = {2025-01-01},
urldate = {2025-01-01},
journal = {Results in Engineering},
volume = {25},
pages = {103776},
abstract = {The inherent ‘black box’ nature of deep learning models has hindered their widespread adoption in certain fields, as they provide limited transparency into the reasoning behind their predictions. In the last years, Explainable Artificial Intelligence (XAI) techniques have proven to be effective not only in prediction itself but also in the extraction of meaningful knowledge from deep learning models by means of feature interpretation. In this study, Local Interpretable Model-agnostic Explanations are applied to the prediction of CTCF binding sites, a common task in the field of genomics. Good prediction performances and inferred explanations are obtained that highlight the most informative features that contribute to predictions such as chromatin accessibility and cis-regulatory elements which align well with previously reported data. This work represents a proof of concept showing that XAI are suitable for the extraction of molecular insights from complex biological problems like CTCF binding prediction.},
keywords = {big data, bioinformatics, deep learning},
pubstate = {published},
tppubtype = {article}
}
The inherent ‘black box’ nature of deep learning models has hindered their widespread adoption in certain fields, as they provide limited transparency into the reasoning behind their predictions. In the last years, Explainable Artificial Intelligence (XAI) techniques have proven to be effective not only in prediction itself but also in the extraction of meaningful knowledge from deep learning models by means of feature interpretation. In this study, Local Interpretable Model-agnostic Explanations are applied to the prediction of CTCF binding sites, a common task in the field of genomics. Good prediction performances and inferred explanations are obtained that highlight the most informative features that contribute to predictions such as chromatin accessibility and cis-regulatory elements which align well with previously reported data. This work represents a proof of concept showing that XAI are suitable for the extraction of molecular insights from complex biological problems like CTCF binding prediction.
2024
H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci
Lecture Notes in Networks and Systems, Springer, vol. 889, 2024, ISBN: 978-3-031-75010-6.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{SOCO2024_part2,
title = {Proceedings of the 19th International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2024) Salamanca, Spain, October 9-11, 2024, Part II},
author = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
editor = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
url = {https://link.springer.com/book/10.1007/978-3-031-75010-6},
doi = {https://doi.org/10.1007/978-3-031-75010-6},
isbn = {978-3-031-75010-6},
year = {2024},
date = {2024-11-19},
volume = {889},
publisher = {Lecture Notes in Networks and Systems, Springer},
series = {Lecture Notes in Networks and Systems},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci
Lecture Notes in Networks and Systems, Springer, vol. 888, 2024, ISBN: 978-3-031-75013-7.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{SOCO2024_part1,
title = {Proceedings of the 19th International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2024) Salamanca, Spain, October 9-11, 2024, Part I},
author = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
editor = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
url = {https://link.springer.com/book/10.1007/978-3-031-75013-7},
doi = {https://doi.org/10.1007/978-3-031-75013-7},
isbn = {978-3-031-75013-7},
year = {2024},
date = {2024-11-15},
volume = {888},
publisher = {Lecture Notes in Networks and Systems, Springer},
series = {Lecture Notes in Networks and Systems},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci
Lecture Notes in Networks and Systems, Springer, vol. 957, 2024, ISBN: 978-3-031-75016-8.
Links | BibTeX | Tags: big data, clustering, education
@proceedings{CISIS-ICEUTE2024,
title = {Proceedings of the International Joint Conference 17th International Conference on Computational Intelligence in Security for Information Systems (CISIS 2024) 15th International Conference on EUropean Transnational Education (ICEUTE 2024). Salamanca, Spain, October 9-11, 2024},
author = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
editor = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
url = {https://link.springer.com/book/10.1007/978-3-031-75016-8},
doi = {https://doi.org/10.1007/978-3-031-75016-8},
isbn = {978-3-031-75016-8},
year = {2024},
date = {2024-11-15},
volume = {957},
publisher = {Lecture Notes in Networks and Systems, Springer},
series = {Lecture Notes in Networks and Systems},
keywords = {big data, clustering, education},
pubstate = {published},
tppubtype = {proceedings}
}
M. J. Jiménez-Navarro and A. R. Troncoso-García and A. Troncoso and F. Martínez-Álvarez and M. Martínez-Ballesteros
Explainable Deep Learning with Embedded Feature Selection for Electricity Demand Forecasting Conference
SST International Conference on Smart Systems and Technologies, 2024.
Abstract | Links | BibTeX | Tags: deep learning, energy, feature selection, XAI
@conference{SST2024,
title = {Explainable Deep Learning with Embedded Feature Selection for Electricity Demand Forecasting},
author = {M. J. Jiménez-Navarro and A. R. Troncoso-García and A. Troncoso and F. Martínez-Álvarez and M. Martínez-Ballesteros},
url = {https://ieeexplore.ieee.org/document/10755283},
doi = {10.1109/SST61991.2024.10755283},
year = {2024},
date = {2024-10-16},
urldate = {2024-10-16},
booktitle = {SST International Conference on Smart Systems and Technologies},
pages = {153-158},
abstract = {Electricity demand forecasting is an important part of the energy industry strategy. Accurate predictions are crucial for maintaining a stable energy supply, planning production, managing distribution, preventing grid overloads, integrating renewable energy sources, and reducing costs and environmental impact. Machine learning and, in particular, deep learning are promising techniques to improve the prediction accuracy of electric demand, but face challenges related to a lack of interpretability due to the “black box” nature of some models. Feature selection methods address these issues by identifying relevant features and simplifying the learning process. This paper aims to explain the most critical lags that impact electric demand forecasting in Spain using the temporal selection layer technique within deep learning models for time series forecasting. This technique transforms a neural network into a model with embedded feature selection, aiming to enhance efficacy and interpretability while reducing computational costs. The results were compared with other methods that incorporate an embedded feature selection mechanism to select the best model. Furthermore, an explainable technique is used to assess the feature importance in the best model over the last year to understand how input features influence electric demand forecasting and provide insights into their contributions and interactions. The results show that our approach improves both the efficacy and interpretability in the context of electric demand forecasting.},
keywords = {deep learning, energy, feature selection, XAI},
pubstate = {published},
tppubtype = {conference}
}
Electricity demand forecasting is an important part of the energy industry strategy. Accurate predictions are crucial for maintaining a stable energy supply, planning production, managing distribution, preventing grid overloads, integrating renewable energy sources, and reducing costs and environmental impact. Machine learning and, in particular, deep learning are promising techniques to improve the prediction accuracy of electric demand, but face challenges related to a lack of interpretability due to the “black box” nature of some models. Feature selection methods address these issues by identifying relevant features and simplifying the learning process. This paper aims to explain the most critical lags that impact electric demand forecasting in Spain using the temporal selection layer technique within deep learning models for time series forecasting. This technique transforms a neural network into a model with embedded feature selection, aiming to enhance efficacy and interpretability while reducing computational costs. The results were compared with other methods that incorporate an embedded feature selection mechanism to select the best model. Furthermore, an explainable technique is used to assess the feature importance in the best model over the last year to understand how input features influence electric demand forecasting and provide insights into their contributions and interactions. The results show that our approach improves both the efficacy and interpretability in the context of electric demand forecasting.
H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci
Springer, vol. 14858, 2024, ISBN: 978-3-031-74185-2.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{HAIS2024_part2,
title = {Proceedings of the 19th International Conference on Hybrid Artificial Intelligent Systems (HAIS 2024) Salamanca, Spain, October 9-11, 2024, Part II},
author = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
editor = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
url = {https://link.springer.com/book/10.1007/978-3-031-74186-9},
doi = {https://doi.org/10.1007/978-3-031-74186-9},
isbn = {978-3-031-74185-2},
year = {2024},
date = {2024-10-10},
volume = {14858},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
A. R. Troncoso-García and M. J. Jiménez-Navarro and M. L. Linares-Barrera and I. S. Brito and F. Martínez-Álvarez and M. Martínez-Ballesteros
Time Series Forecasting in Agriculture: Explainable Deep Learning with Lagged Feature Selection Conference
SOCO 19th International Conference on Soft Computing Models in Industrial and Environmental Applications, Lecture Notes in Networks and Systems 2024.
Links | BibTeX | Tags: deep learning, forecasting, time series, XAI
@conference{SOCO24_Troncoso,
title = {Time Series Forecasting in Agriculture: Explainable Deep Learning with Lagged Feature Selection},
author = {A. R. Troncoso-García and M. J. Jiménez-Navarro and M. L. Linares-Barrera and I. S. Brito and F. Martínez-Álvarez and M. Martínez-Ballesteros},
url = {https://link.springer.com/chapter/10.1007/978-3-031-75013-7_14},
doi = {https://doi.org/10.1007/978-3-031-75013-7_14},
year = {2024},
date = {2024-10-10},
booktitle = {SOCO 19th International Conference on Soft Computing Models in Industrial and Environmental Applications},
pages = {139-149},
series = {Lecture Notes in Networks and Systems},
keywords = {deep learning, forecasting, time series, XAI},
pubstate = {published},
tppubtype = {conference}
}
H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci
Springer, vol. 14857, 2024, ISBN: 978-3-031-74182-1.
Links | BibTeX | Tags: big data, clustering, deep learning, IoT
@proceedings{HAIS2024_part1,
title = {Proceedings of the 19th International Conference on Hybrid Artificial Intelligent Systems (HAIS 2024) Salamanca, Spain, October 9-11, 2024, Part I},
author = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
editor = {H. Quintián and E. Corchado and A. Troncoso and H. Pérez-García and E. Jove-Pérez and J. L. Calvo-Rolle and F. J. Martínez de Pisón and P. García-Bringas and F. Martínez-Álvarez and Á. Herrero and P. Fosci},
url = {https://link.springer.com/book/10.1007/978-3-031-74183-8},
doi = {https://doi.org/10.1007/978-3-031-74183-8},
isbn = {978-3-031-74182-1},
year = {2024},
date = {2024-10-09},
volume = {14857},
publisher = {Springer},
series = {Lecture Notes in Artificial Intelligence},
keywords = {big data, clustering, deep learning, IoT},
pubstate = {published},
tppubtype = {proceedings}
}
F. Rodríguez-Díaz and A. M. Chacón-Maldonado and A. R. Troncoso-García and G. Asencio-Cortés
Explainable Olive grove and Grapevine pest forecasting through machine learning-based classification and regression Journal Article
In: Results in Engineering, vol. 24, pp. 103058, 2024.
Abstract | Links | BibTeX | Tags: deep learning, feature selection, time series, XAI
@article{RODRIGUEZ24,
title = {Explainable Olive grove and Grapevine pest forecasting through machine learning-based classification and regression},
author = {F. Rodríguez-Díaz and A. M. Chacón-Maldonado and A. R. Troncoso-García and G. Asencio-Cortés},
url = {https://www.sciencedirect.com/science/article/pii/S2590123024013136},
doi = {https://doi.org/10.1016/j.rineng.2024.103058},
year = {2024},
date = {2024-09-09},
urldate = {2024-09-09},
journal = {Results in Engineering},
volume = {24},
pages = {103058},
abstract = {Pests significantly impact agricultural productivity, making early detection crucial for maximizing yields. This paper explores the use of machine learning models to predict olive fly and red spider mite infestations in Andalusia. Four datasets on crop phenology, pest populations, and damage levels were used, with models developed using the Python package H20, which focuses on interpretability through SHAP values and ICE plots. The results showed high precision in predicting pest outbreaks, particularly for the olive fly, with minimal differences between models using feature selection. In the vineyard dataset, the selection of characteristics improved the performance of the model by reducing the MAE and increasing R2. Explainability techniques identified solar radiation and wind direction as key factors in olive fly predictions, while past pest occurrences and wind velocity were influential for red spider mites, providing farmers with actionable insights for timely pest control.},
keywords = {deep learning, feature selection, time series, XAI},
pubstate = {published},
tppubtype = {article}
}
Pests significantly impact agricultural productivity, making early detection crucial for maximizing yields. This paper explores the use of machine learning models to predict olive fly and red spider mite infestations in Andalusia. Four datasets on crop phenology, pest populations, and damage levels were used, with models developed using the Python package H20, which focuses on interpretability through SHAP values and ICE plots. The results showed high precision in predicting pest outbreaks, particularly for the olive fly, with minimal differences between models using feature selection. In the vineyard dataset, the selection of characteristics improved the performance of the model by reducing the MAE and increasing R2. Explainability techniques identified solar radiation and wind direction as key factors in olive fly predictions, while past pest occurrences and wind velocity were influential for red spider mites, providing farmers with actionable insights for timely pest control.
O. S. Mazari and A. Sebaa and F. Martínez-Álvarez
Space-Time Clustering of Seismicity in Algeria Conference
CSA 6th Conference on Computing Systems and Applications, Lecture Notes in Networks and Systems 2024.
Abstract | Links | BibTeX | Tags: clustering, natural disasters
@conference{CSA24_Mazari,
title = {Space-Time Clustering of Seismicity in Algeria},
author = {O. S. Mazari and A. Sebaa and F. Martínez-Álvarez},
url = {https://link.springer.com/chapter/10.1007/978-3-031-71848-9_36},
doi = {https://doi.org/10.1007/978-3-031-71848-9_36},
year = {2024},
date = {2024-08-08},
booktitle = {CSA 6th Conference on Computing Systems and Applications},
pages = {396–405},
series = {Lecture Notes in Networks and Systems},
abstract = {Each year, earthquakes pose a significant threat to human life, attributed to their sudden and unpredictable nature. Over time, a heightened awareness of this phenomenon has driven increased attention from researchers and experts. This paper seeks to demonstrate the applicability of the k-means algorithm to seismic data, focusing on the identification of seismic zones in Algeria. Initially, we conducted a comprehensive review of existing literature on clustering seismic data, revealing an unexplored niche in the context of Algeria’s seismicity. Subsequently, we introduce our dataset comprising 5876 seismic events. A detailed explanation of the k-means algorithm is provided, with a breakdown of each parameter. Visualization of our findings, including determining the optimal value for k using Elbow and Silhouette scores, is presented and thoroughly discussed. In conclusion, we identify and delineate the seismic zones in Algeria, highlighting the four most critical regions encapsulating these zones. This study contributes to a better understanding of seismic patterns in Algeria, potentially aiding in the development of more effective earthquake preparedness and mitigation strategies.},
keywords = {clustering, natural disasters},
pubstate = {published},
tppubtype = {conference}
}
Each year, earthquakes pose a significant threat to human life, attributed to their sudden and unpredictable nature. Over time, a heightened awareness of this phenomenon has driven increased attention from researchers and experts. This paper seeks to demonstrate the applicability of the k-means algorithm to seismic data, focusing on the identification of seismic zones in Algeria. Initially, we conducted a comprehensive review of existing literature on clustering seismic data, revealing an unexplored niche in the context of Algeria’s seismicity. Subsequently, we introduce our dataset comprising 5876 seismic events. A detailed explanation of the k-means algorithm is provided, with a breakdown of each parameter. Visualization of our findings, including determining the optimal value for k using Elbow and Silhouette scores, is presented and thoroughly discussed. In conclusion, we identify and delineate the seismic zones in Algeria, highlighting the four most critical regions encapsulating these zones. This study contributes to a better understanding of seismic patterns in Algeria, potentially aiding in the development of more effective earthquake preparedness and mitigation strategies.