An Interval-Varying Floating Car Data Clustering Approach Based onSelf-Organizing Maps

摘要

Floating car data (FCD) are characterized by high dimensions and largedata size with complex spatiotemporal relationships. This paper proposes a timeinterval-varying scheme for FCD dimensional reduction for clustering analysis,which can capture the main fluctuation closer to the original data. Aself-organizing map is used as the clustering algorithm. Some interesting andhidden relationships between links can be found in the clustering results. Themethod proposed in this paper proved acceptable for FCD clustering, which isuseful for road correlation analysis of the urban road network.