Knowledge Graphs-Based Halal Food Provider Information Search System at Surabaya

In Indonesia, approximately 10% of products possess halal certification. Consequently, there is a need for a solution, specifically a system capable of retrieving information from halal food providers and generating recommendations. We crawl the BPJPH API. Subsequently, the ranking of food providers is determined using the Node Similarity algorithm and graph embedding utilizing the Node2vec and Fast Random Projection algorithms in conjunction with the K-Nearest Neighbors algorithm. A simple search engine application was then developed using StreamLit, an open-source Python-based application framework designed to facilitate the creation of web applications in the fields of machine learning and data science. Application development also incorporates SQLiteStudio for database creation. Following the search engine application design process, an evaluation stage was conducted using two metrics: Mean Reciprocal Rank (MRR), a metric to assess each process that gives potential answers to a query arranged in order of their likelihood of being correct, and Normalized Discounted Cumulative Gain (NDCG), a metric that evaluates performance of a ranking list by penalizing incorrect sequence in the list. Overall, the performances of systems that use Node Similarity and K-Nearest Neighbors with Fast Random Projection algorithms are generally better than the performance of system that uses K-Nearest Neighbors with Node2Vec Algorithms.