Rail Transit Station Classification Based on Spectral Clustering Algorithm

The classification of urban rail transit stations can provide basis for urban managers to plan industrial layout, rail transit station operation plan, rail transit passenger flow forecast, etc. First, this paper standardizes the AFC credit card data of Hangzhou rail transit, land use data around the station, bus connection data around the station, and road network data around the station. After principal component analysis, three main factors are obtained. Then, three common factors extracted as clustering variables are employed, and k-means++ method and spectral clustering are adopted to cluster the track stations, respectively. By comparing the clustering index Calinski-Harabasz index, it is reached that spectral clustering is better. The results show that urban rail transit stations can be divided into four categories by spectral clustering: residential, commercial business, employment, and transport hub. Based on the results of site classification, we can analyze and summarize the universal law of passenger flow of various types of sites to prepare for the subsequent site passenger flow forecast.