Mining Trip Attractive Areas Using Large-Scale Taxi Trajectory Data

摘要

With the rapid development of location aware computing technologies, the research of data mining for spatial-temporal trajectories has attracted many scholars. As an important data source, through the GPS receivers taxis provide researchers with massive trajectory data, which are featured by their high quality, good continuity and wide distribution, making it suitable for travel pattern mining. In this paper, we study on urban residents' trip attractive areas with trajectory data of about 10000 taxis in Chongqing city, to discover the characteristics of urban traffic distribution to serve for the construction of modern intelligent traffic system. Firstly, a framework of trajectory preprocessing, including data cleaning, extraction for the taxi passenger pick-up points and drop-off points, is presented to reduce the noise and redundancy in raw trajectory data. Secondly, in the light of the shortcomings of the existing mining methods for attractive areas, such as high complexity and poor scalability, we propose a grid density based clustering algorithm to mine the trip attractive areas in different periods. In the experiments, we take a performance comparison between our algorithm and DBSCAN algorithm, and discuss how to determine the optimal parameters of our algorithm. Then, we verify the practicability of our algorithm with real data.