Diffusion-based location-aware recommender systems

The recommendation is now part of our daily life. As the years pass by, companies collect more and more information about the users of their platforms. One question which could arise is: are the data collected useful for better predictions? In this paper, we investigate the performance impact on adding geographical positions on the performance of the prediction of users' behavior using an existing diffusion-based recommender system. We show how we can improve the accuracy of the diffusion algorithm using the geographical position of users. The accuracy of the improved algorithm is compared with the state of art similar recommender algorithms. Moreover, we design a general framework to infer the position location of users based on the position of their activities. (paper: interdisciplinary statistical mechanics)