Discovery of Important Location from Massive Trajectory Data Based on Mediation Matrix

Abstract

Analyzing large volume of trajectory data plays an important role in understanding user behaviors and providing personalized recommendations. However, existing work faces many challenges in important location discovery processing speed and accuracy. This paper proposes a general computing framework to improve the accuracy of occupational and residential location detection in cellular network. An important location discovery module and an index structure is included, which improves the efficiency and accuracy. A mining algorithm MMA (Matrix base Mining Algorithm) is proposed, which improves the accuracy of user important location. Experimental evaluation shows that the proposed algorithm has higher accuracy and efficiency in real environment.