Published in Electric Power Systems Research, 2026
The paper proposes a multivariate time series classification method with imbalance-aware preprocessing and ensemble learning to improve Wind Power Ramp Event forecasting, achieving over 85% accuracy and 88% weighted F1 score on real-world wind power data.
Published in Transportation, 2025
This paper proposes a novel stochastic, simulation-based calibration method that uses only traffic count data to improve the accuracy, scalability, and real-time applicability of traffic models, demonstrating a 16% accuracy gain over state-of-the-art methods in a Brussels case study.
Published in IEEE Access, 2024
This paper reviews the literature on single-objective constrained Bayesian optimization, classifying methods by metamodels, acquisition functions, and constraint handling, while outlining real-world applications, limitations, and future research directions.
Published in Engineering with computers, 2024
This study applies advanced learning and Bayesian optimization to efficiently identify Pareto-optimal adhesive bonding process parameters under multiple noisy objectives and strict experimental constraints, reducing the need for costly physical testing.
Published in Artificial Intelligence Review, 2022
This article systematically surveys multi-objective hyperparameter optimization methods from 2014–2020, categorizing algorithms, evaluating comparison metrics, and highlighting future research directions.
Published in Expert Systems with Applications, 2022
This paper demonstrates that Long Short-term Cognitive Networks (LSTCNs) provide faster and more accurate online forecasting of windmill time series than traditional RNN-based models, making them well-suited for real-time windmill monitoring and maintenance applications.
Published in Renewable energy, 2022
This paper introduces two fuzzy concept–based methods for wind turbine health monitoring, enabling interpretable detection of performance degradation by analyzing changes and drifts in power production under varying environmental conditions.
Published in Computación y Sistemas, 2021
This paper proposes two new diversity measures for classifier ensembles, based on coverage and similarity, to improve ensemble accuracy and provides experimental analysis showing their effectiveness and correlation with existing diversity metrics.
Published in Investigacion operacional, 2021
This paper proposes a modified bagging algorithm that integrates diverse learning algorithms and optimizations to improve classifier ensembles, demonstrating superior performance over state-of-the-art methods and validating its effectiveness in real biochemical applications.
Published in Revista Cubana de Informática Médica, 2019
This paper applies neurofuzzy systems to diagnose hypertension risk in children, showing that the NSLV algorithm generates effective diagnostic rules from clinical data to support early detection.
Published in Investigación Operacional, 2017
This paper presents an Ant Colony Optimization-based approach to construct efficient and diverse ensemble of classifiers, optimizing classifier selection to enhance accuracy while maintaining model simplicity.
Published in Computacion y Sistemas, 2016
In this paper we present the different diversity measures that exist in the literature to decide if a set of classifiers is diverse, aspect that is very important in the creation of ensembles.
Published in Investigación Operacional, 2015
This paper proposes a modified genetic algorithm that uses diversity measures to select the most diverse and accurate combination of classifiers, demonstrating its effectiveness through applications in two different domains.
Published in Revista Cubana de Ciencias Informáticas, 2015
This study addresses the prediction of true and false splice sites in DNA sequences by applying and comparing multiple machine learning algorithms in WEKA, concluding that Bayesian methods provide the best classification performance based on measures such as true positive rate and ROC area.
Published in Book: Experiencias en la modelación de la toma de decisiones en la salud humana, medio ambiente y desarrollo humano, 2015
This study proposes a genetic algorithm based multiclassifier that combines 18 base classifiers and 13 diversity measures to improve the early diagnosis of hypertension in children aged 10 to 12 years, achieving greater accuracy than individual classifiers.
Published in Communications in Computer and Information Science, 2023
This study leverages Gaussian Process and Logistic Regression models within a Bayesian optimization framework to handle multi-objective, constrained, and uncertain adhesive bonding optimization, achieving efficient solutions with minimal experimental effort.
Published in Communications in Computer and Information Science, 2022
This paper proposes a hybrid multi-objective hyperparameter optimization method that combines Tree-structured Parzen Estimators with Gaussian Process Regression under heterogeneous noise, improving performance under uncertainty compared to stand-alone approaches.
Published in Proceedings in Engineering Mechanics, 2022
This paper demonstrates that Bayesian optimization with Gaussian process models can effectively support adhesive bonding process optimization, achieving similar joint strength as expert-driven methods with up to 40% less budget and reduced production costs, while ensuring robust and durable bonds.