<![CDATA[Hadith is the second primary source of Islamic law. This study aims to classify the translations of Sahih al-Bukhari Hadith into three thematic categories simultaneously, namely recommendations, prohibitions, and information, using a multi-label classification approach with the Backpropagation Neural Network (BPNN) method. The dataset consists of 7,000 labeled instances. The preprocessing stage includes case folding and text cleaning without stopword removal or stemming to preserve the original meaning of the text. The novelty of this study lies in the comprehensive exploration of various model configurations, including variations in hidden layer architecture, learning rate values, and n-gram features, to identify the optimal network architecture. Feature extraction was performed using Term Frequency–Inverse Document Frequency (TF-IDF) with n-gram configurations of (1,1) and (1,2) and a maximum of 7,000 features. Three variations of the BPNN architecture (H1, H2, and H3) were evaluated using 10-fold cross-validation. The experimental results show that the simplest architecture, consisting of a single hidden layer with 64 neurons, a learning rate of 0.001, and an n-gram configuration of (1,1), achieved the best performance, with macro accuracy of 91.43%, F1-score of 89.59%, and Hamming loss of 0.0857.]]>
Copyrights © 2026
