A Dynamic Statistical Model for Metro Smart Card Data
This paper studies the analysis issue of metro passenger flows.We constructs a dynamic statistical model for passengers'tap-in and tap-out times.This model does not requrietrain schedules.An EM algorithm is proposed to solve the maximum likelihood estimates of unknown parameters in our model.With passengers'tap-in and tap-out times,the proposed model can be applied to infer several indices of metro systems,such as train schedules,passengers'boarding probabilities,distributions of egress times,and distributions of travel times.This point indicates that we provide a solution to the inferential problem of passenger flows in peak hours,which is not extensively discussed in the literature.Simulation results show that the proposed EM algorithm can yield accurate estimates of unknown parameters in the model.Furthermore,we apply our model to analyze a real data set from Line 6 in Beijing metro.Several quantities that describe passenger flow features in both peak and off-peak hours are presented.It is concluded that the time is a significant factor to the boarding probability and distribution of egress time.Based on the proposed model,we construct dynamic prediction intervals for passengers'tap-out times.The results on a test set indicate that the actual coverage rate is consistent with the nominal level,which also show the effectiveness of the proposed model.
万方数据
维普
