Diffusion Kernel Based Mobility Prediction for Wireless Users

Abstract

The issue about mobility prediction has attracted many researchers from diverse disciplines due to its critical role in various applications. In this paper, we propose a prediction model based on diffusion kernel and demonstrate its effectiveness using a large-scale real-world Call Detail Record(CDR) data set. First, we describe the diffusion kernel based mobility prediction model in detail. Then we implement the next place prediction for the users in the CDR data set with the proposed prediction model and compared its prediction performance with Markov-based models. The comparison shows that the diffusion-based model performs better. Besides, the mobility prediction theory based on entropy is utilized to measure the predictability of users in the data set. Based on the predictability result, we analyze the correlation between prediction accuracy of the proposed model and predictability. We validates the effectiveness of predictability and also reveal that the prediction accuracy varies even for users with the same entropy, which means that people with the same entropy might have different prediction accuracy.