Feature selection using binary grey wolf optimizer with elite-based crossover for Arabic text classification

Text classification is one of the challenging computational tasks in machine learning community due to the increased amounts of natural language text documents available in the electronic forms. In this process, feature selection (FS) is an essential phase because thousands of possible feature sets may be considered in text classification. This paper proposes an enhanced binary grey wolf optimizer (GWO) within a wrapper FS approach to tackle Arabic text classification problems. The proposed binary GWO is utilized to play the role of a wrapper-based feature selection technique. The performance of the proposed method using different learning models, including decision trees,K-nearest neighbour, Naive Bayes, and SVM classifiers, are investigated. Three Arabic public datasets, namely Alwatan, Akhbar-Alkhaleej, and Al-jazeera-News, are utilized to evaluate the efficacy of different BGWO-based wrapper methods. Results and analysis show that SVM-based feature selection technique with the proposed binary GWO optimizer with elite-based crossover scheme has enhanced efficacy in dealing with Arabic text classification problems compared to other peers.