Abstract

Nowadays, scientists can collect and analyze massive mobile data generated by various sensors and applications of smart phones. smart phones have become an important platform for the understanding of social activities, such as community detection, social dynamics and influence. It is extremely important to store and retrieve mobile data efficiently for various data mining tasks. In this paper, we propose Mobile Data Warehouse (<em>MobileDW</em>) model which is based on <em>GraphChi</em>, a system designed for large-scale graph computation on one PC. We propose multi-shard data structure and Time-based Parallel Sliding Windows (<em>TPSW</em>) to store Social data such as call logs and SMS. We further propose Mobile Index (<em>MIndex</em>) structure and Mobile Position Compression Algorithm (<em>MPCA</em>) to warehouse Position data such as GPS, Bluetooth etc. The <em>MIndex</em> structure can compress <em>Position</em> data significantly. The data compression process is based on the following observations: (1) The position of the individual users within a certain period of time often unchanged. (2) A crowd of people tend to move and stay together. Experimental results demonstrate the effectiveness and efficiency of Mobile Data Warehouse.