A Multi-swarm Particle Swarm Optimization with an Adaptive Regrouping Strategy for Feature Selection
Chenye Qiu · 2020
Feature selection is an important data pre-processing step which aims at removing irrelevant and redundant features. It can enhance the performance of the learning model and reduce the computational complexity. Particle swarm optimization (PSO) has been widely used in feature selection due to its rapid convergence, but it suffers from premature convergence when solving high-dimensional feature selection problem. Hence, this paper proposes a multi-swarm PSO to overcome the existing drawback. The multi-swarm is capable of maintaining good population diversity and shows strong global exploration ability. An adaptive regrouping strategy is proposed to promote the information exchange among the sub-swarms and speed up the convergence speed. Furthermore, a local search operator is hybridized to improve the local exploitation ability. The proposed MSPSO-A is compared with five PSO based wrappers on 11 UCI datasets. Experimental results demonstrate the proposed algorithm can effectively improve the classification accuracy and reduce the number of features. Furthermore, two statistical tests are performed to show the superiority of the proposed method over other methods is significant.