dinner4two

dinner4two is a recommendation system that helps couples on their first dates to find the ideal restaurant that will please both. In this project, we leveraged Yelp's dataset to build our MVP, initially focused on restaurants in Austin, TX. The MVP consists of a web-based front-end that takes input from both users and a recommender engine that provides a list of restaurants that takes into consideration both users' preferences. Our app allows both users to select pictures of suggested locations they may like, in order to collect initial features and address the cold start problem. For our recommendation system, we evaluated 3 models: a simple baseline model that used cosine similarity, a more complex model based on matrix factorization, and a third model using a Graph Convolutional Neural Network. The latter provided the best accuracy and was the selected model for this MVP.