A Spatiotemporal Cluster Method for Trajectory Data

Abstract

As positioning and communication technologies become more widespread, the production of large amounts of different types of trajectory data and the extraction of useful information from mass trajectory data have emerged as hot issues in data mining. This paper presents a trajectory data processing method featuring simple operation, high precision, and strong practicability. For low-precision trajectory data that are discrete but contain time information, a clustering algorithm is proposed to extract information from such data. The algorithm can detect a point of interest (POI) in trajectory data by setting space and time thresholds. Trajectory data collected from a taxi using a global positioning system in Kunming, China, are used as experimental data. We conduct an experiment to detect a POI in the collected trajectory data and carry out a visual analysis of these special positions. The experimental results show the effectiveness of the algorithm, which can in addition compress trajectory data.