Analysis on Commuted Passenger Characteristics and Impact of Guangzhou Rail Transit by Machine Learning Based on Data Fusion

Commuter traffic is the core subject of urban rail transit passenger flow. Influenced by multiple factors such as the expansion of the network scale, the application of diversified transportation methods, and the introduction of electric bicycles, the characteristics of commuter traffic in Guangzhou's Metro have shown new trends. Therefore, this paper conducts a systematic analysis of four commuting characteristics-portrait, spatial distribution, gathering-dispersion, and connectivity of Metro in Guangzhou by integrating data from AFC, Baidu, mobile signaling, and the bus-card. It also utilizes machine learning methods to explore influencing factors of characteristics and their coupling relationships. The research results indicate that the main body of Guangzhou's rail transit commuters consists of white-collar workers whose age are young and middle-aged with medium to high consumption habits and middle to low-income. Approximately 70% of their residences and 80% of their workplaces are within 1 km of a subway station. The connecting distance to the subway stations out the city central area is about 3.3 times that of the internal areas, and the connecting distance at workplaces increases at a faster rate as the geographical location extends outward compared to residences. Furthermore, rail commuting exhibits a significant centripetal tendency, with the degree of centrifugal aggregation of positions within line-service-area and average station spacing of line having a notable positive impact on the directional imbalance coefficient and average travel distance. The conversion rate of jobs and population within 800m around stations which is less than 10% are 76.67% and 89.08% of the total number of stations respectively. Additionally, there is a certain negative correlation between the average travel distance of commuter traffic at stations and the connectivity of the stations.