Short-Term Passenger Flow Forecast in Urban Rail Transit Based on Enhanced K-Nearest Neighbor Approach

Passenger flow forecasting is important for transit service planning and operational management. Many models and techniques have been developed to address this issue. In order to improve the accuracy of short-term passenger flow forecasting, an enhanced K-nearest neighbor method is proposed in this paper. The method considers the trend factor and time interval factor of passenger flow in the stage of state vector design and distance measurement, and then avoids the risks of fewness of the evaluation criterion of the original method in the matching process. Based on smart card data from the automatic fare collection system of the subway in Beijing, we designed an experiment to test the ability of the new method and three models. The test results show that the improvement scheme has better performance of forecasting comparing with BP neural network model (BPNN), SARIMA model and the original KNN method.