A trust and relevance-based Point-Of-Interest recommendations method with inaccessible user location
Gladysheva Ekaterina, Ivan Derevitskii, Severiukhina Oksana · Procedia Computer Science · 2020
Point of interest recommender services are widely studied and applied in various areas. In most cases, these services are based on user location data. However, in some cases, the user’s location may not be available. For example, if the user has closed access to this data on a mobile device or in the case of using a web service that does not have accurate location information. Nevertheless, it is necessary to provide the user with the most relevant POIs. In this paper, we propose a modification of the collaborative filtering method based on the trust and relevance of user experience to solve this problem. We compare the quality of ranking using the proposed method with a set of classical interpreted recommender algorithms, such as content algorithms, and matrix decomposition algorithms. Public catering enterprises (bars, cafes, restaurants) in St. Petersburg were used as data example. The feedback data of 90814 customers and 4256 catering places in St. Petersburg were collected from “Tripadvisor” and “Restoclub” services. To estimate the quality of the results, ranking metric MAP@K was used. Experimental results showed that the proposed method is 1.104 times more efficient (from the point of view of the MAP@K metric) than existing approaches.