Bus Station Location Selection Method Based on DBSCAN-DPC Clustering Algorithm

Bus station selection requires cluster analysis of passenger travel data, and the passenger flow distribution data is relatively scattered and usually does not have spherical distribution characteristics. DBSCAN clustering was selected as the basic algorithm, to solve the problem that DBSCAN algorithm cannot obtain cluster centers, density peak clustering (DPC) algorithm is introduced to search cluster centers. DBSCAN-DPC algorithm can recognize outliers of passenger travel data, and uses a decision graph heuristic method to find the point with the large local density and offset distance as the clustering center. Through pre-processing and cluster analysis of 42 bus lines, 1079 bus stops and 1048575 IC card data in Wangjing area of Beijing, the results show that the algorithm has better clustering effect on non-spherical data, reduces the influence of outliers on the results, and can effectively identify the cluster center in the area with dense passenger flow. The peak density point in the cluster is taken as the station location, which improves the station coverage in the area with dense passenger flow, and conforms to the demand-oriented characteristics of bus.