Mobile user Location Prediction Based on user Classification and Markov Model

Abstract

Individual's location has play a very important role in mobile network, predicting user's next location in advance has a great effect on optimize network resource and improve user experience. The prediction of mobile user location has become a hot topic of research, but the accuracy of prediction is still a bottleneck. Instead of directly predicting an individual's next location based on the historical data of all users, in this paper, we propose a location prediction method based on the mobile user classification. Firstly, we propose the user classification criteria and the corresponding classification algorithm according to their daily behavior patterns. Secondly, we propose a trajectory prediction scheme for each user cluster based on their characteristics. Using the real cellular network data, simulation results show that the prediction accuracy of the proposed method is up to 66%, which is around 8% higher than the traditional second-order Markov model.