.png){kind=icon}
.png){kind=icon}
Return
Fast-TRACLUS: An Optimized Trajectory Clustering Algorithm for Large-Scale Datasets
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-032-08462-0_23.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Trajectory clustering is a crucial task in various domains, from urban mobility to wildlife monitoring, as it enables the discovery of representative movement patterns and identification of anomalous behaviours. TRA-CLUS is a well-established algorithm for trajectory segmentation and clustering; however, its high computational cost limits its applicability in large-scale or real-time scenarios. This study introduces FAST-TRACLUS, an optimised version of TRA-CLUS that significantly reduces execution time without compromising cluster quality. The optimisation involves efficient data structures, parallel processing and geometric simplifications. A comprehensive experimental study using real-world trajectory datasets demonstrates that FAST-TRACLUS achieves an average runtime improvement of 27.4% over the original implementation while maintaining equivalent clustering accuracy. Furthermore, a modular clustering framework is presented, supporting five alternative algorithms-DBSCAN, OPTICS, HDBSCAN, Spectral Clustering, and Agglomerative Clustering-to enable comparative evaluation and flexibility in analysis. The results highlight the effectiveness of OPTICS as a balanced choice for both execution time and quality across various datasets, while also showcasing the strengths and limitations of other clustering methods in terms of handling noise, data distribution, and computational complexity. This study bridges the gap between algorithmic efficiency and practical usability, offering researchers and practitioners a versatile and high-performance tool for trajectory data analysis.
Keywords:
TRACLUS
Large-scale datasets
Computational efficiency
Journal
H
IF:
0
Papers:
28
Citations:
0
