An Improved Binary Particle Swarm Optimization with Complementary Distribution Strategy for Feature Selection

Feature selection is a preprocessing technique with great importance in the fields of data analysis, information retrieval processing, pattern classification, and data mining applications. It process constitutes a commonly encountered problem of global combinatorial optimization. This process reduces the number of features by removing irrelevant, noisy, and redundant data, thus resulting in acceptable classification accuracy. This paper presents a novel optimization algorithm called complementary binary particle swarm optimization (CBPSO), in which using complementary distribution strategy to improve the search capability of binary particle swarm optimization (BPSO) by facilitates global exploration and local exploitation via complementary particles and original particles, respectively. This complementary approach is introduction new particles (“complementary particles”) into the search space, which generated by half of the whole particles selected randomly, and replaces the selected particles when the fitness of the global best particle has not improved for a number of consecutive iterations. In this study, the K-nearest neighbor (K-NN) method with leave-one-out crossvalidation (LOOCV) was used to evaluate the quality of the solutions. CBPSO was applied and compared to ten classification problems taken from the literature. Experimental results show that CBPSO simplifies the feature selection process effectively, and either obtains higher classification accuracy or uses fewer features than BPSO and other feature selection methods.