Topic Classification of Arabic Text Using Majority Voting

Arabic, being the original language of the Quran, plays an essential part in effectively communicating its profound teachings. The Al-Quran is a sacred scripture that serves as a comprehensive guide for followers of Islam. Al-Quran covers various topics such as faith, deeds, science, and more. Therefore, to facilitate the understanding of the verses of Al-Quran, it is important to categorize the topics of each verse. The model developed in this study represents a multi-label classification system built using the majority voting approach. This approach was chosen because it generally provides better model performance than just one basic model. Subsequently, the model's performance is measured using two types of measures hamming loss and micro F1-score. Based on the experiments, the best performance is achieved using a combination of Random Forest, SVM, and Adaptive Boosting models. This combination achieved 0.5143 for the micro F1-Score and 0.0878 for the hamming loss. Additionally, the majority voting method improves model performance compared to the models used individually.