关键词: OPTICS算法, 轨迹聚类, 空间聚类方法

Abstract: With the development of mobile location technology and the popularity of mobile location devices, analysis of moving object trajectory data has become a hot topic in the field of spatial data mining. Based on the spatial clustering of taxi trajectory data, it can find the hot path and trend of taxi movements. Based on the traditional OPTICS（Ordering Points To Identify the Clustering Structure） algorithm by trajectory data, TR-OPTICS algorithm（Trajectory OPTICS） is proposed according to the characteristics of mass trajectory. This method uses passenger taxi trajectory as the research object, selects the feature points and divides the trajectory by MDL（Minimum Description Length） method, calculates the horizontal distance, vertical distance, angle distance of sub-trajectory to measure the similarity. In the clustering algorithm, it uses outer rectangle as search band of the core-trajectory, redefines the reach-distance and core-distance, and uses adjacency list to reduce the complexity of the algorithm. Through the Nanjing city taxi trajectory analysis, it gets the taxi hot trajectory clusters, and through repeated experiments contrast with the traditional OPTICS algorithm, DBSCAN（Density-Based Spatial Clustering of Applications with Noise） algorithm, this paper’s TR-OPTICS algorithm is better than the other two algorithms, trajectory clustering results are mainly in center road of Nanjing city, such as Central Road, Bridge South Road, Beijing East Road, Zhongshan East Road, Zhongshan North Road, Jianning Road, Ruijin Road, Itakura Street, Maigaoqiao Road. The results show that this paper’s TR-OPTICS algorithm improves the efficiency of the algorithm and improves the accuracy of clustering results.

Key words: Ordering Points To Identify the Clustering Structure（OPTICS） algorithm, trajectory clustering, spatial clustering method