基于时空约束密度聚类的职住地识别方法

为了从移动终端位置数据中精准识别居民职住地，提出了一种基于时空约束密度聚类的职住地识别方法。首先，利用基于K-means的DBSCAN(density-based spatial clustering of applications with noise)时空驻点聚类过程将居民多天的原始轨迹点分成不同的时空驻点簇；然后，利用基于速度阈值的停留点簇和移动点簇识别过程将居民的每一个时空驻点簇区分为停留点簇或移动点簇；接着，利用基于K近距离的DBSCAN重要停留点聚类过程将居民的停留点分成不同的重要停留点簇；最后，利用基于KD-tree优化的KNN(K-nearest neighbor)职住地识别过程将居民的每个重要停留点识别为工作地、居住地、职住同一区域或兴趣地点区域。实验结果表明，该方法的每个过程都是合理有效的，并且最终的职住地识别效果要优于时间阈值法、累加时间法和信息熵法。

Home-work location identification method based on spatiotemporal constrained density clustering

To accurately identify residential home-work locations from mobile terminal location data, this paper proposed a home-work location identification method based on spatiotemporal constrained density clustering. Firstly, the method used a K-Means based DBSCAN spatiotemporal stationary point clustering process to divide the original trajectory points of many days for each resident into different spatiotemporal stagnation point clusters. Then, it used a recognition process of residence point cluster and moving point cluster based on velocity threshold to recognize every spatiotemporal stationary point cluster of each resident as a stationary point cluster or a moving point cluster by. After that, it used a DBSCAN important residence point clustering process based on K-nearest distance to divide the residence points of each resident into different important residence point clusters. Finally, it used a KNN home-work location identification process optimized by KD-tree to identify every important residence point of each resident as a home location, a work location, a home-work location or an interest location. The experimental results show that each process of this method is reasonable and effective, and the final recognition effect of home-work locations is better than the time threshold method, the cumulative time method and the information entropy method.