User Location Forecasting Based on Collective Preferences

Abstract

With the proliferation of mobile devices and the huge variety of sensors they incorporate, it is possible to register the user location on the move. Based on historical records, it is feasible to predict user location in space or space and time. Studies show that user mobility patterns have a high degree of repetition and this regularity has been exploited to forecast the next location of the user. Furthermore, proposals have been made to forecast user location in space and time; in particular, we present a spatio-temporal prediction model that we developed to forecast user location in a medium-term with good accuracy results. After explaining how collaborative filtering (CF) works, we explore the feasibility of using collective preferences to avoid missing POIs and therefore increase the prediction accuracy. To test the performance of the method based on CF, we compare our spatio-temporal prediction model with and without using the method based on CF.