Crowd location forecasting at points of interest

Predicting the location of a mobile user in the near future can be used for a large number of user-centered ubiquitous applications. This can be extended to crowdcentered applications if a large number of users is included. In this paper we present a spatio-temporal prediction approach to forecast user location in a medium-term period. Our approach is based on the hypothesis that users exhibit a different mobility pattern for each day of the week. Once factored out this weekly pattern, user mobility among points of interest is postulated to be markovian. We trained a hidden Markov model to forecast user mobility and evaluated our approach using a public dataset. The experimental results show that our approach is effective considering a time period of up to seven hours. We obtained an accuracy of up to 81.75 % for a period of 30 minutes, and 66.25 % considering 7 hours.