Deep Learning-Based Passenger Flow Hotspot Prediction for Urban Cab Trips

The taxi-cab has been regarded as one of the influential public traffic modes with the characteristics of “point-to-point,” which can provide a high-quality data set for analyzing traffic hotspots. In this study, we use the GPS data of cabs in Hejin city, Shanxi Province, to extract the spatio-temporal distribution of traffic hotspots. Afterwards, the deep learning model, LSTM, is developed to predict traffic hotpots by judging the next point of individual trajectories under different weather and period conditions. Results show that the proposed approach has a good prediction accuracy of 97%. The findings can help to guide dispatching cabs, relieve urban traffic pressure, and reduce passengers’ waiting time.