A new method of railway passenger flow forecasting based on spatio-temporal data mining

By analyzing the limitation of current passenger flow forecasting approach, This work presents a new approach to forecast the railway passenger flow based on spatio-temporal data mining. The approach first forecasts the time sequence of the target object using statistical principles, then figures out the spatial influence of neighbor objects using a neural network, and finally combines the two forecasting results using linear regression. The method is used in the forecast of railway passenger flow during the spring festival period of 2004. Comparing with the existing approaches that do not consider the spatial influence, our approach has better forecast accuracy.