Short-time Forecast Of High-speed Railway Passenger Flow Based On Ensemble Empirical Mode Decomposition-gray Support Vector Machine

In recent years, China’s high-speed railway undertakings have developed rapidly. China has become the country in the world with the fastest growing, the most complete system technology, the strongest integration capabilities, the longest operating mileage, highest operating speed, largest under construction of high-speed railway. However, the development of high-speed railway in China is inseparable from the basic research. Passenger flow is the basis and foundation to build the high-speed rail. Reasonable organization of passenger flow is the key to play a high-speed rail benefits. Therefore, the analysis of high-speed railway passenger flow, accurate and reasonable short-term high-speed rail passenger flow forecasting methods has important theoretical value and practical significance.Firstly, the paper reviewed the theories and methods of high-speed railway passenger flow forecast from three angles of passenger formation mechanism, passenger behavior characteristics and combination forecasting. This laid the theoretical basis of Short-term forecast model. Secondly, the paper analyzed the volatility characteristics of the high-speed rail passenger from cycle fluctuations, vibration, energy and mode-aliasing by statistics and time series mining. The paper analyzed high-speed railway passenger flow fluctuation signal by using the EMD. Thirdly, constructed the EEMD-GSVM short-term forecasting model which fusion of gray generation and support vector machine with the concept of "isolation" forecast and improved EEMD model. Then studied modal reconstruction and parameter optimization program.Fourthly, take the Wuhan-guangzhou high-speed railway as an example and utilizing the models the issue of short-term high-speed railway passenger flow forecast has been analyzed.The high accuracy forecast results descriptioned the better adaptability of EEMD-GSVM model.Finally, summarized the research results and put forward some questions for future researeh.