Enhancing Hadith Classification using Statistical and Semantic Feature Fusion and Dimension Reduction

Recent global interest in Hadith literature has encouraged efforts to enhance the efficiency of Hadith classification in Arabic. This study investigates the impact of different text representation models on Hadith classification. We introduce a novel model combining two text representation techniques: Term Frequency-Inverse Document Frequency (TF-IDF) and word embedding followed by dimension reduction using Principal Component Analysis (PCA). This approach allows us to emphasize word importance and capture the semantic meaning of each word within the dataset. Our dataset includes 834 hadiths from Sahih Al-Bukhari, representing five distinct categories. The model has been evaluated using various machine-learning algorithms, and the results show that the Stochastic Gradient Descent (SGD) algorithm achieved the highest F1-score ($86.99 \%$) followed by the Multi-Layer Perceptron (MLP) algorithm with (81.53%). These findings indicate the potential of our approach to significantly improve the categorization of Hadith in the Arabic language, which can have a profound impact on research in Islamic studies.