Konferensartikel

The uncertainty of estimating the railway passenger flow in advance may disrupt the passenger operation and management (e.g., passenger evacuation planning, seat allocation, and train timetable programming). In order to proactively improve the service quality and efficiency of the railway system, the short-term passenger flow prediction technique is vital in the field of operation and management system. Utilizing the deep learning library-keras, the study develops a Long short-term memory neural network (LSTM NN) to predict the short-term high-speed rail (HSR) passenger flow. Processing the raw data, we first construct the passenger flow sequences as the input (output) variables. Then the gird search and cross validation techniques are applied to optimize the LSTM NN parameters. At last we utilize the data provided by Shanghai railway administration of China as the case study. Through a comparison with other representative methods, including Auto-Regressive Integrated Moving Average (ARIMA), Back Propagation Neural Network (BPNN), and Support Vector Machine Regression (SVR), results suggest that the proposed LSTM NN can generate great potentials for accurate passenger flow predictions.

Short-term passenger flow prediction, High-speed rail, Long short-term memory neural network, Grid search, Cross validation