Real-time Trajectory Movement Pattern Detection Based On Apache Flink

With the prevalence of GPS devices,trajectory data is produced and collected from a wide spectrum of applications in an unprecedented scale.The analysis of trajectory data is of profound value in both scientific research and business services.Among the analytics tasks,trajectory co-movement pattern detection is an important type.So far,many kinds of trajectory co-movement patterns have been studied in literature and they have been proved to be valuable in real-life.However,with the emerging on-line services,the traditional off-line algorithm can no longer meet the needs of real time analysis.Recently,some research findings about real time trajectory data analysis come out,but all of these works only focus on simple data queries like range query or nearest neighbor query.The problem of real time trajectory co-movement pattern detection yet has not been thoroughly studied.Motivated by this situation,we present a formal definition of real time trajectory co-movement pattern detection and propose an algorithm called ICPE(indexed clustering and pattern enumeration)based on Apache Flink.Besides,we prove the effectiveness and efficiency of ICPE with a series of experiments.