Food Recommendation System Based on Collaborative Filtering and Taste Profiling

With restaurant menus becoming vast and dish names complicated, we present a recommendation system to facilitate ordering food at restaurants. The design approach is three-pronged and comprises the following options: (I) recommendations for you: menu items that are custom selected for a user through collaborative filtering based on the rating of dishes previously ordered and a targeted model that selects items based on a detailed comparison of reviews, such as any mention of ingredients, cooking technique, etc., where such reviews are available, (ii) similar items: recommendations that provide alternatives based on the ingredients and similar method of preparation of items previously selected and (iii) frequently bought together: recommendations that target increasing the ‘revenue per order’ based on association rule mining using the apriori algorithm. All these models have been incorporated into an existing food-ordering platform – ‘Instead’ – developed by the team and beta-tested at multiple locations. The recommendation system presented in this work will enable a user to walk into any restaurant and order an item consistent with their taste palette.