HCSA: A Novel Feature Selection Method based on Improved Crow Search Algorithm

Abstract

As a simple and efficient swarm intelligence optimization algorithm, Crow Search Algorithm (CSA) has been widely used in many practical fields. But the traditional Crow Search Algorithm also faces the risk of falling into a local optimum when dealing with complex optimization problems, and there exists blind search phenomenon in the optimization process. In order to improve the algorithm’s ability to get rid of local optima and apply it to select the best feature subset, this paper proposes a wrapper feature selection algorithm (HCSA) based on an improved Crow Search Algorithm. Firstly, try to use different methods to initialize the population to obtain the initial population with better quality. Then, the concept of “leader” is introduced to guide the optimization to reduce blind search. The position of the leader is adjusted by the strategy of Salp Swarm Algorithm (SSA). In order to enhance the diversity, chaotic map is added in the position updating equation. In addition, the nonlinear parameter updating strategy is used to dynamically adjust the important parameter AP in the original algorithm during the iteration to achieve an effective balance between exploration and exploitation. The experimental results indicate that HCSA shows good performance in solving the feature selection problem.