Location Prediction Based on Variable-order Markov Model and User's Spatio-temporal Rule

Ying Xia, Yu Gong, Xu Zhang, Hae-Young Bae

Abstract: Location based services are widely used and can bring convenience to people's life. Improving the accuracy of location prediction is important for the service quality. We propose a location predication method VMSTPM which combining the variable order Markov model and user's spatio-temporal rule. According to user's prediction request, this method firstly abstracts active grids from user's historical trajectory and extracts the stay points, and then the variable-order Markov model is established to predict the location according to the matching situation between current trajectory and historical one. For the possible zero frequency problem, the spatio-temporal rule of the users' activities is obtained by analyzing the historical trajectory. Experiments show that this method can improve the validity and accuracy of location prediction, and it is more suitable for location prediction with fewer trajectory data.