an effective feature selection with social mimic optimization algorithm

Hundreds of variables in data lead to data with very high dimensions, allowing many feature selection methods to be developed. the purpose of feature selection in machine learning, pattern recognition, and data mining is to choose features that will enhance learning performance. the aim of this paper is to use the binary version of the social mimic optimization (smo) algorithm as binary social mimic optimization (bsmo) for feature selection. the combined fitness function is chosen because of its three main objectives: reducing classification error, balancing sensitivity and specificity, and reducing the number of selected features. the proposed method is compared with several optimization methods, including binary genetic algorithms (bga) and particle swarm optimization (bpso), as well as with binary atom search optimization (baso). the results of the evaluation using five uci datasets show that the proposed method is superior to others for solving optimization problems.