An Integrated and Dynamic Commuter Flow Forecasting System for Railways

Abstract

Uncertain and instable passenger flow in Urban Metro Transit is a growing concern in the recent rail transport system. It is vital to forecast the passenger flow, in order to provide a reliable daily operation and management. Short-term forecasting has become the most important component for an efficient rail management system. Existing literatures on passenger flow forecasting is based on extreme kernel approach that learns and forecasts signals with different frequencies. These approaches are not able to train and remember over a long time due to issues of backpropagated error. By addressing this problem, holt-winters forecasting algorithm is used. Experimental discussion shows that the holt algorithm provides better efficiency based on metrics including accuracy and F-measure.