Temporal Markov Chain Location Prediction

Abstract

With the rapid development of social mobile networks, it is highly desirable to be able to accurately predict the next check-in location. Most methods of social mobile networks are mainly based on vast user check-in historic records to analyze the feature of users’ behaviors. The Markov Model is a frequently used method in location prediction problem. However, due to the diversity and sparsity of the social mobile network check-in data, tradition Markov Model location prediction, cannot take full advantages of potential information existed in plentiful check-in data. Then, we propose a prediction algorithm, temporal Markov Model (TMM) based on time characteristic. This algorithm relies on Markov Model to make a primary prediction of user checkin position. Then, utilizing the temporal character modifies the forecast result. We have conducted extensive experiments in two real datasets, and the results demonstrate the superiority of TMM over traditional Markov Models.