Halal Food Products Recommendation Based on Knowledge Graphs and Machine Learning

Halal cuisine is an essential requirement for every Muslim. However, the number of halal-certified products are less than number of non-halal-certified products. Therefore, we determine whether a non-halal-certified food product is similar to halal-certified products based on the relationship of shared ingredients using machine learning and knowledge graphs. This study compares the product from an online grocery website with the Halal Linked Open Data (LOD) dataset using the Naive Bayes, KNN and Random Forest methods. Features extraction using several graph algorithms: Common Neighbors, Preferential Attachment, Total Neighbors, Label Propagation and Louvain. The results of the performance calculation are assessed from accuracy, precision, recall and F1-Score. Random Forest performance surpasses other machine learning performance. We found that performing link prediction can be done with high accuracy rate by using the traditional machine learning method and can be optimized further by performing hyperparameter tuning with large datasets.