TrajSummary: Summarizing Trips of Individuals

Abstract

City-scale mobility data of people is recorded by multiple location sensing applications. User trajectory summarization is one of the important applications that provides the movement pattern summary of users, which can be used in applications such as city transportation planning, hyper-targeted advertising. We propose TrajSummary: a system for summarizing and quantifying mobility patterns of an individual. We evaluate three novel approaches to cluster user trips - THRESH, L-AWARE and MODE-EST. We show that our techniques provide significantly superior user summary clusters than generic trajectory clustering mechanisms such as SWARM and TRA-CLUS. Our approach is faster; we perform 5x faster than techniques using Dynamic Time Warping based approaches.