TMFS-MTS: Toolbox for metaheuristic feature selection in multivariate time series

Abstract

Feature selection in multivariate time series is a key challenge in modern data analysis, as high-dimensional data often include temporal dependencies and irrelevant features degrading classifier performance. To address these issues, a comparison with existing approaches is essential. Therefore, this work introduces the Toolbox for Metaheuristic Feature Selection in Multivariate Time Series (TMFS-MTS), providing an environment for feature selection and metaheuristic evaluation. It supports diverse fitness measures and advanced visualizations, including convergence curves, feature count tracking, runtime analysis, Wilcoxon tests, and 2D embeddings. Implemented in MATLAB, TMFS-MTS offers a standardized framework for advancing research in multivariate time series feature selection.