Doble Grado en Ingeniería Informática - Tecnologías Informáticas y Matemáticas
URI permanente para esta colecciónhttps://hdl.handle.net/11441/176210
Examinar
Envíos recientes
Trabajo Final de Grado (TFG) GAUFS: Genetic Algorithm for Unsupervised Feature Selection for Clustering(2026-06-24) de-la-Torre González, Salvador; Sánchez López, José Enrique; Carranza García, Manuel; Lenguajes y Sistemas InformáticosThis Bachelor’s thesis presents GAUFS (Genetic Algorithm-based Unsupervised Feature Selection for Clustering), a wrapper methodology for clustering in machine learning that jointly addresses feature selection and estimation of the number of clusters in a fully unsupervised setting. It employs a genetic algorithm to explore candidate solutions that encode feature subsets and cluster configurations, guided by internal clustering validation criteria. The evolutionary search acts as an exploratory mechanism, and its outcomes are analyzed within the GAUFS pipeline to obtain the final solution. This thesis extends the original GAUFS work presented in the scientific paper [1] by providing an expanded methodological description, a more detailed experimental analysis, and a comprehensive implementation overview. The framework has been refactored and released as an open-source Python package available on PyPI, improving usability, modularity and reproducibility. The method is evaluated on 30 synthetic and 30 real-world datasets against seven established state-of-the-art approaches, achieving competitive performance and consistent improvements in clustering quality while simultaneously estimating the number of clusters. In addition, two synthetic data generators are proposed to enable a controlled evaluation of the relevance of the feature and the separability of the cluster. Overall, this thesis provides a comprehensive methodological extension and a fully usable software tool for unsupervised feature selection in clustering, supporting further research and practical adoption.