Abstract: In order to effectively guide traffic dispatching, prevent and deal with rail transit emergencies, the curve fitting method was used to mine the trend characteristics of passenger flow time series. Based on the analysis of time distribution law of passenger flow and the exploration of data, the time series modeling and optimization of passenger flow in 36 subway stations in Beijing in a single day were carried out by global fitting, artificial piecewise fitting and automatic piecewise fitting, and a comparison in both the fitting results and R² index was made between the three methods. The research shows that the piecewise fitting uses the local function to model the dynamic process of passenger flow change, which can better approximate the actual situation than the global fitting, that in the case of piecewise fitting, the automatic piecewise strategy is adopted, which can avoid the subjectivity in artificially determining piecewise points, and achieve optimal optimization, further improving the efficiency and accuracy of prediction analysis.

Key words: urban rail transit, passenger flow, time series, trend characteristics, piecewise fitting