A MRT Daily Passenger Flow Prediction Model with Different Combinations of Influential Factors

Nowadays mass passenger flow phenomena at different degrees appear in many big cities with large-scale transit network, and those phenomena tend to be regular and complicated. As the most popular public transport in Taipei, Mass Rapid Transit (MRT) is a quite effective traffic tool to relieve the pressure of passenger congestion, especially in rush hours. In this paper, a MRT passenger flow prediction model with deep neural network (DNN) is proposed. The various combinations of influenced factors including historical passenger flow, temporal factors, directional factor and holiday factor have been trained as the inputs by using the dataset collected from Taipei Main stations. The experiments results revealed that the modeling approach and the input influenced factors are both important elements to affect the performances of the passenger flow prediction model.