Efficient Arabic Text Classification Using Feature Selection Techniques and Genetic Algorithm

The rapid growth of the Internet in recent years has been accompanied by a huge increase in the number of data being shared at every moment, especially text data. This huge amount of text data requires finding new techniques to be able to deal with the data. One form of these texts is the Arabic texts which poses a big challenge to researchers due to the highly derivative nature of Arabic texts. In this study, we first aim to create an effective classification model for Arabic texts, where we will start by creating a huge high-quality dataset consisting of different versions collected by web scraping techniques. Then, we apply term weighting (TW) and feature selection (FS) procedures to be able to classify our collected Arabic dataset and propose an approach that is based on genetic algorithm (GA) to ensure that the best subset of features is selected with high quality. The results prove that our proposed approach can select up to 1% of the original feature size with high efficiency. The results provide detailed information about the result of each classification technique and TW schemes, and runtime needed to obtain the best results to aid in future research aimed at developing Arabic text classification techniques.