A Dynamic Tasking-Based Evolutionary Algorithm for Bi-Objective Feature Selection DOI Creative Commons
Hang Xu

Mathematics, Journal Year: 2024, Volume and Issue: 12(10), P. 1431 - 1431

Published: May 7, 2024

Feature selection in classification is a complex optimization problem that cannot be solved polynomial time. Bi-objective feature selection, aiming to minimize both selected features and errors, challenging due the conflict between objectives, while one of most effective ways tackle this use multi-objective evolutionary algorithms. However, very few these have ever reflected an multi-tasking framework, despite implicit parallelism offered by population-based search characteristic. In paper, dynamic multi-tasking-based algorithm (termed DTEA) proposed for handling bi-objective classification, which not only suitable datasets with relatively lower dimensionality features, but also higher features. The role influence on were studied, tasking mechanism self-adaptively assign multiple tasks intermittently analyzing population behaviors. efficacy DTEA tested 20 compared seven state-of-the-art A component contribution analysis was conducted comparing its three variants. empirical results show dynamic-tasking works efficiently enables outperform other algorithms terms classification.

Language: Английский

Algorithm Initialization: Categories and Assessment DOI
Abdul Halim, Swagatam Das, Idris Ismail

et al.

Emergence, complexity and computation, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 100

Published: Jan. 1, 2024

Language: Английский

Citations

1
A Multi-Task Decomposition-Based Evolutionary Algorithm for Tackling High-Dimensional Bi-Objective Feature Selection DOI Creative Commons
Hang Xu,

Chaohui Huang,

Jianbing Lin

et al.

Mathematics, Journal Year: 2024, Volume and Issue: 12(8), P. 1178 - 1178

Published: April 14, 2024

Evolutionary algorithms have been widely applied for solving multi-objective optimization problems, while the feature selection in classification can also be treated as a discrete bi-objective problem if attempting to minimize both error and ratio of selected features. However, traditional evolutionary (MOEAs) may drawbacks tackling large-scale selection, due curse dimensionality decision space. Therefore, this paper, we concentrated on designing an multi-task decomposition-based algorithm (abbreviated MTDEA), especially handling high-dimensional classification. To more specific, multiple subpopulations related different tasks are separately initialized then adaptively merged into single integrated population during evolution. Moreover, ideal points these dynamically adjusted every generation, order achieve search preferences directions. In experiments, proposed MTDEA was compared with seven state-of-the-art MOEAs 20 datasets terms three performance indicators, along using comprehensive Wilcoxon Friedman tests. It found that performed best most datasets, significantly better ability promising efficiency.

Language: Английский

Citations

1
A Dynamic Tasking-Based Evolutionary Algorithm for Bi-Objective Feature Selection DOI Creative Commons
Hang Xu

Mathematics, Journal Year: 2024, Volume and Issue: 12(10), P. 1431 - 1431

Published: May 7, 2024

Feature selection in classification is a complex optimization problem that cannot be solved polynomial time. Bi-objective feature selection, aiming to minimize both selected features and errors, challenging due the conflict between objectives, while one of most effective ways tackle this use multi-objective evolutionary algorithms. However, very few these have ever reflected an multi-tasking framework, despite implicit parallelism offered by population-based search characteristic. In paper, dynamic multi-tasking-based algorithm (termed DTEA) proposed for handling bi-objective classification, which not only suitable datasets with relatively lower dimensionality features, but also higher features. The role influence on were studied, tasking mechanism self-adaptively assign multiple tasks intermittently analyzing population behaviors. efficacy DTEA tested 20 compared seven state-of-the-art A component contribution analysis was conducted comparing its three variants. empirical results show dynamic-tasking works efficiently enables outperform other algorithms terms classification.

Language: Английский

Citations

0