Recommender systems have become a very important tool in everyday life, supporting users in purchasing goods online and finding entertainment. Researchers who develop culinary recommender systems use collaborative filtering-based recommender systems that usually only provide overall ratings without details. Culinary tourism requires more detailed recommendations, such as food flavor, ambiance, service, price, and distance. Currently, most collaborative filtering-based culinary recommendation systems primarily consider rating data. However, an individual's preference for an item is often influenced by the features contained within that item. To overcome this problem, we develop an efficient and scalable culinary recommendation system with the Alternating Least Squares (ALS) algorithm for large datasets, using data of dining places in Yogyakarta that includes various information such as restaurant name, location, food price, rating, and service type. We use Alternating Least Squares (ALS) to build a culinary recommender system in the Yogyakarta area. The advantage of the ALS method is that it has high scalability and efficiency in handling large and sparse datasets, making it possible to provide accurate recommendations. For evaluation, we compared the performance of ALS with Singular Value Decomposition (SVD) using i.e. RMSE and Precision metrics. Results show ALS is superior with RMSE 0.2518 and Precision 0.4017 compared to Singular Value Decomposition (SVD) which has RMSE 0.2571 and Precision 0.3890. ALS is able to provide more accurate and relevant culinary recommendations, improving the user experience in finding places to eat that match their preferences.
Keywords—culinary recommender system, collaborative filtering, alternating least squares