Abstract
Feature selection is the process of removing duplicated and unimportant features from the dataset in order to enhance the learning algorithm and data mining. In this paper, a hybrid form of optimization algorithm as a wrapper feature selection model is used to reduce the number of features while trying to improve the classification accuracy. This hybrid algorithm used a binary variant of both Particle Swarm Optimization (PSO) and Spotted Hyena Optimization Algorithm (SHO), which is called BSHPSO. The proposed approach has been tested on five high-dimensional low-instances medical datasets from UCI. The results showed superior performance for the BSHPSO over the binary versions of both PSO and SHO algorithms.
Keywords: Hybrid Algorithm, Binary Optimization Algorithm, Transfer Function, Feature Selection, Particle Optimization Algorithm, Spotted Hyena Optimization