Feature selection, construction and search space reduction based on genetic programming for high-dimensional datasets.
Saved in:
| Title: | Feature selection, construction and search space reduction based on genetic programming for high-dimensional datasets. |
|---|---|
| Authors: | Herrera-Sánchez, David1 (AUTHOR) hersan19@hotmail.es, Mezura-Montes, Efrén1 (AUTHOR) emezura@uv.mx, Acosta-Mesa, Héctor-Gabriel1 (AUTHOR) heacosta@uv.mx, Márquez-Grajales, Aldo1 (AUTHOR) li.aldomg@gmail.com |
| Source: | Neural Computing & Applications. Nov2025, Vol. 37 Issue 33, p27557-27574. 18p. |
| Subjects: | Feature selection, Genetic programming, Heuristic, Classification |
| Abstract: | Nowadays, analyzing high-dimensional data containing thousands of features is utmost importance. As a result, the classification task has become challenging due to the computational cost necessary to achieve competitive accuracy levels. However, feature selection and construction have emerged as solutions to this problem during preprocessing. The primary objective of these techniques is to identify the most pertinent features from the dataset, which diminishes the run-time during the learning process and improves the classification accuracy. Genetic programming can efficiently handle these preprocessing techniques automatically and cohesively by utilizing tree representation. This paper outlines a method for narrowing the search space in high-dimensional datasets through a simple heuristic that eliminates irrelevant features during the search process. This approach reduces the search area, ensuring that only the most representative features for the chosen classifier are selected. The paper uses twelve datasets. The first subset ranges from three hundred to seven thousand features. The second subset ranges from ten thousand to twenty-four thousand features, and the third subset only contains a dataset with more than one hundred thirty thousand features. The findings indicate that the search area can be decreased by over 50% without impacting the classification accuracy. Furthermore, the convergence of the optimal solutions using the heuristic is improved regardless of not using it, achieving values of accuracy higher than 95% of classification. [ABSTRACT FROM AUTHOR] |
| Copyright of Neural Computing & Applications is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Nowadays, analyzing high-dimensional data containing thousands of features is utmost importance. As a result, the classification task has become challenging due to the computational cost necessary to achieve competitive accuracy levels. However, feature selection and construction have emerged as solutions to this problem during preprocessing. The primary objective of these techniques is to identify the most pertinent features from the dataset, which diminishes the run-time during the learning process and improves the classification accuracy. Genetic programming can efficiently handle these preprocessing techniques automatically and cohesively by utilizing tree representation. This paper outlines a method for narrowing the search space in high-dimensional datasets through a simple heuristic that eliminates irrelevant features during the search process. This approach reduces the search area, ensuring that only the most representative features for the chosen classifier are selected. The paper uses twelve datasets. The first subset ranges from three hundred to seven thousand features. The second subset ranges from ten thousand to twenty-four thousand features, and the third subset only contains a dataset with more than one hundred thirty thousand features. The findings indicate that the search area can be decreased by over 50% without impacting the classification accuracy. Furthermore, the convergence of the optimal solutions using the heuristic is improved regardless of not using it, achieving values of accuracy higher than 95% of classification. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 09410643 |
| DOI: | 10.1007/s00521-024-10567-7 |
