Optimization and Empirical Evaluation of Passenger Leaving Station in Bus Based on Gradient Charge Swiping Card Data

Abstract:

This paper proposed a fusion analysis model for leaving station identification， based on the 5-day continuous （24 h） bus travel data， bus travel GPS data， and bus stop data of a certain city， in combination with the travel chain method and random forest network. In the model， the GPS data and bus stop data were first matched to determine the bus arrival information at different moments， then the passenger boarding location， travel frequency， activity scope， land use around the leaving station， and the probability of getting off the bus were applied as inputs to identify the passenger leaving station. The final calculation rate is improved to 100%， and the full sample efficiency rate reaches 76.2%. Compared with the existing methods based on the bus travel chain， the recognition efficiency is improved by 37%.