Feature selection is an essential task in machine learning and data mining that involves identifying a subset of relevant features from a larger set. This paper proposes a novel technique for unsupervised feature selection based on a Neural Network in conjunction with an evolutionary algorithm. The proposed method aims to extract subsets of the most discriminative and relevant features from high-dimensional data, which can be eventually used for efficient and accurate machine learning. An evolutionary algorithm is employed to generate the feature subsets, and the goodness of a feature subset is evaluated through the ability of a neural network to reconstruct the whole original input space by mean squared error minimization (in an auto-encoder fashion). Experimental results demonstrate the effectiveness of the proposed approach in finding relevant feature subsets for successive learning tasks, achieving better classification and regression accuracy compared to state-of-the-art feature selection methods.

U-FLEX: Unsupervised Feature Learning with Evolutionary eXploration / Bellarmino, Nicolo’; Cantoro, Riccardo; Squillero, Giovanni. - ELETTRONICO. - 14505:(2024), pp. 364-378. (Intervento presentato al convegno The 9th International Conference on Machine Learning, Optimization, and Data Science (LOD 2023) tenutosi a Grasmere, Lake District, England (UK) nel September 22 – 26, 2023) [10.1007/978-3-031-53969-5_27].

U-FLEX: Unsupervised Feature Learning with Evolutionary eXploration

Nicolo’ Bellarmino;Riccardo Cantoro;Giovanni Squillero
2024

Abstract

Feature selection is an essential task in machine learning and data mining that involves identifying a subset of relevant features from a larger set. This paper proposes a novel technique for unsupervised feature selection based on a Neural Network in conjunction with an evolutionary algorithm. The proposed method aims to extract subsets of the most discriminative and relevant features from high-dimensional data, which can be eventually used for efficient and accurate machine learning. An evolutionary algorithm is employed to generate the feature subsets, and the goodness of a feature subset is evaluated through the ability of a neural network to reconstruct the whole original input space by mean squared error minimization (in an auto-encoder fashion). Experimental results demonstrate the effectiveness of the proposed approach in finding relevant feature subsets for successive learning tasks, achieving better classification and regression accuracy compared to state-of-the-art feature selection methods.
2024
978-3-031-53968-8
File in questo prodotto:
File Dimensione Formato  
2023_LOD_Genetic_Feature_Selection (2).pdf

embargo fino al 16/02/2025

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 430.26 kB
Formato Adobe PDF
430.26 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
978-3-031-53969-5_27.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 1.35 MB
Formato Adobe PDF
1.35 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2981873