Cycle Periodic Behavior Detection and Sports Place Extraction Using Crowdsourced Running Trace Data

Abstract

Crowdsourcing trace data mining plays an important role in behavior pattern mining, place sensing, etc. This paper proposes a new method to automatically detect cycle periodic behavior and extract outdoor sports place from running tracks. First, the cycle periodic behavior is modeled using movement parameters. Second, based on the features of cycle periodic pattern, the trajectory distance matrix search algorithm is presented to detect periodic behavior and extract periodic tracks. Last, the sports place information is extracted by Delaunay triangulation and reverse geocoding method from collective cycle periodic tracks. Experiments were conducted using one month smartphone app running traces in Beijing, and the results show that the proposed method can more effectively identify cycle periodic pattern compared to the Apriori method, and it can efficiently extract sports place information.