Application of cluster analysis with unsupervised learning to dockless shared bicycle flow control and dispatching

Abstract

After dockless shared bicycles are introduced on a large scale in the city, supply and demand problems involving different areas often emerge. This study applied cluster analysis with unsupervised learning to shared bicycle flow control and dispatching, and employed artificial intelligence to extract real, full-scale transportation rules from open data. First, this study proposes a model of the shared bicycle control system and an incentive mechanism for the reverse flow of bicycles based on threshold values. Then, we use the kernel density spatial clustering method to perform partitioning, grading, and incentives of check-out and check-in points’ density in the area, furthermore, adopt the DBSCAN clustering method to establish dispersal and dispatching strategies. This study uses Shanghai Open Data in modeling, verification, and used Rstudio software to produce visualized interactive graphics for demonstration.