Feature Selection Using Hybrid Binary Grey Wolf Optimizer for Arabic Text Classification

Abstract

Feature selection in Arabic text is a challenging task due to the complex and rich nature of Arabic. The feature selection requires solution quality, stability, convergence speed, and the ability to find the global optimal. This study proposes a feature selection method using the Hybrid Binary Gray Wolf Optimizer (HBGWO) for Arabic text classification. The HBGWO method combines the local search capabilities or exploratory of the BGWO and the search capabilities around the best solutions or exploits of the PSO. HBGWO method also combines SCA's capabilities in finding global solutions. The data set used Arabic text from islambook.com, which consists of five Hadith books. The books selected five classes: Tauhid, Prayer, Zakat, Fasting, and Hajj. The results showed that the BGWO-PSO-SCA feature selection method with the fitness function search and classification method using SVM could perform better on Arabic text classification problems. BGWO-PSO with fitness function and the classification method using SVM (C=1.0) gives a high accuracy value of 76.37% compared to without feature selection. The BGWO-PSO-SCA feature selection method provides an accuracy value of 88.08%. This accuracy value is higher than the BGWO-PSO feature selection and other feature selection methods.