Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China.
Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing 100124, China.
Sensors (Basel). 2023 Apr 5;23(7):3749. doi: 10.3390/s23073749.
The widespread adoption of intelligent devices has led to the generation of vast amounts of Global Positioning System (GPS) trajectory data. One of the significant challenges in this domain is to accurately identify stopping points from GPS trajectory data. Traditional clustering methods have proven ineffective in accurately identifying non-stopping points caused by trailing or round trips. To address this issue, this paper proposes a novel density peak clustering algorithm based on coherence distance, incorporating temporal and entropy constraints, referred to as the two-step DPCC-TE. The proposed algorithm introduces a coherence index to integrate spatial and temporal features, and imposes temporal and entropy constraints on the clusters to mitigate local density increase caused by slow-moving points and back-and-forth movements. Moreover, to address the issue of interactions between subclusters after one-step clustering, a two-step clustering algorithm is proposed based on the DPCC-TE algorithm. Experimental results demonstrate that the proposed two-step clustering algorithm outperforms the DBSCAN-TE and one-step DPCC-TE methods, and achieves an accuracy of 95.49% in identifying stopping points.
智能设备的广泛应用导致了大量全球定位系统（GPS）轨迹数据的产生。在这个领域，一个重大的挑战是如何从 GPS 轨迹数据中准确地识别出停止点。传统的聚类方法在准确识别由于拖尾或往返而导致的非停止点方面效果不佳。针对这个问题，本文提出了一种新颖的基于一致性距离的密度峰值聚类算法，结合了时间和熵约束，称为两步 DPCC-TE。所提出的算法引入了一个一致性指数来整合空间和时间特征，并对聚类施加时间和熵约束，以减轻由于缓慢移动点和往返运动导致的局部密度增加。此外，为了解决一步聚类后子聚类之间的相互作用问题，提出了一种基于 DPCC-TE 算法的两步聚类算法。实验结果表明，所提出的两步聚类算法优于 DBSCAN-TE 和一步 DPCC-TE 方法，在识别停止点方面的准确率达到 95.49%。
