Feature selection for user motion pattern recognition in mobile networks
Mobility is a challenging issues in mobile networks which significantly impacts the performance of a variety of network protocols. Understanding user motion behavior can improve the performance of mobile network protocols in different aspects. Therefore, we consider how to analyze the motion behavior of mobile nodes in various environments. In this paper we propose a few mobility metrics useful for analysis of individual, collective and geographical behavior of mobile nodes and a simple supervised mobility pattern recognition method which is able to classify mobility traces into different mobility model classes using our proposed mobility metrics. The remaining challenge is to find appropriate features, extracted from mobility traces, which are able to distinguish between different motion behaviors with maximum accuracy and minimum overhead. We propose a method to find the most suitable feature sets for classification of different mobility traces with different motion behaviors. Simulation results show significant performance of our proposed mobility pattern recognition method using appropriate mobility metrics and feature sets.