Automation of Halal Food Classification Using Bidirectional Long Short-term Memory on Ingredients List

The global demand for halal food products continues to increase, particularly among Muslim consumers, necessitating an efficient and accurate halal classification system. This study proposes a deep learning-based automatic classification approach using Bidirectional Long Short-Term Memory (BiLSTM) to determine the halal or haram status of a product based on its ingredient list. The system utilizes comprehensive text preprocessing techniques such as normalization, stopword removal, and dictionary-based term mapping. Word representations are converted into dense semantic vectors using word embeddings such as Word2Vec and GloVe. A BiLSTM model is used to capture bidirectional contextual relationships in ingredient sequences, thereby enhancing semantic understanding. Testing results on a dataset of 3,979 samples show that the proposed model achieves a classification accuracy of 99.75%, outperforming traditional machine learning methods such as Naive Bayes and SVM. The system is proven effective in handling ingredient ambiguity and context-based classification, and has potential for real-world applications such as mobile-based halal scanners. Future research can adopt attention mechanisms and transform-based models to improve performance and interpretability.