OPTIMIZING HADITH CLASSIFICATION WITH NEURAL NETWORKS: A STUDY ON BUKHARI AND MUSLIM TEXTS

The Bukhari and Muslim hadith collections encompass a total of 7008 hadith sentences, but it is not immediately clear which of these hadiths fall into the categories of prohibitions or orders. To enhance understanding and accessibility for readers, this study focuses on classifying these hadiths through a systematic process. The classification involves several key stages: Text Pre-processing, pre-processing the raw text data to clean and normalize (Stemming, Stopword Removal and Tokenization), Word vector features are extracted to capture the semantic relationships and contextual meanings of the words, then processed into a neural network model based on a multilayer perceptron (MLP) architecture (Model Architecture, Training and Optimization). The approach leverages the strength of neural networks, particularly through the use of multiple layers and feature extraction via word vectors, which significantly contributes to the accuracy of the classification process. The results of the study is very good, with a high accuracy rate of 97.72% achieved by employing a model with two layers and 256 neurons