A new Markov-based mobility prediction algorithm for mobile networks

Abstract

Mobility prediction is an important solution to enable seamless handovers in cellular networks and the mobility trace is the main information used to perform it. However, using solely this information makes the prediction process difficult when the mobile user is new in the network, that is, when its mobility trace is poor. In this paper, we investigate a Markov-based prediction model which focuses on new mobile users behaviour prediction. In order to assess our approach, we use data sets of a real cellular network in a major US urban area. The efficiency of the prediction model relies on both the ability of the model to predict successfully the next move of a mobile user and its ability to perform such a prediction in a short delay. Comparing our approach with previous solutions, we show that our solution outperforms in all cases the previous solutions and essentially succeeds to make better predictions for new mobile users. © 2010 Springer-Verlag.