Research On Usability Analysis And Improvement Of The Prediction Models Of Passenger Flow Based On Correlation Measurement

Passenger flow forecasting is an important foundation for the railway network planning, line and station design, railway operation and other work. However, it is a difficult work to achieve an accurate prediction of a large passenger flow. Domestic and foreign scholars have done a lot about traffic forecast, however, when do the traffic forecast, there are some problems such as how to select a prediction model and so on.Aiming at the problems that exist in passenger flow forecasting, the paper mainly completed the following work.Firstly, the paper analyzes the relationship between autocorrelation characteristics of passenger flow data and the application effect of prediction models. Focusing on the relationship between correlation characteristics of passenger flow data and the application effect of prediction models, the paper chooses exponential smoothing method, adaptive adjustment method, seasonal adjustment method, and pick-up method as comparison models, and uses the practical data from railway ticket sale and reservation system to research the contrast of prediction effect, experiments of passenger data autocorrelation characteristic analysis, and the relationship between the experiment results of passenger data autocorrelation characteristics analysis and the experiment results of the contrast of prediction effect.Then, based on the above research, the paper puts forward a method to select or construct a suitable passenger flow forecasting model by analyzing the autocorrelation characteristics of historical data. And by using the actual data, the paper verifies the validation of the method.Finally, the paper puts forward a model that uses the correlation of passenger flow historical data to calculate the similarity of ticketing curve. And the model is applied to the Three Stage Model which Tsung-Hsien Tsa proposed, then the paper makes a contrast about prediction effect between the new model and the former model and verifies the validation of the improvement.