Machine Learning Approach to Optimal Restaurant Location Search

Business placement tasks often suffer from the lack of real shop (or restaurant) success data. Even in the presence of success data in the current location, it is difficult to estimate the restaurant's success for a wide range of possible locations. In this work, several approaches are proposed for assessing restaurant success with open data. We have developed a method for recommending locations of commercial establishments based on machine learning from open data of St. Petersburg's restaurants. As a result of the experiment on the data of St. Petersburg restaurants, the proposed method showed higher-ranking quality than heuristic methods, and the method showed stability for metric changing.