Research On Prediction Model Of Railway Passenger Volume Based On Data Feature

Railway transportation plays an extremely important role in raising the speed of the traffic transportation,logistics transportation and promoting national economic development which is one of the main body of long-distance transport market.Railway passenger traffic can reflect the demand levels of the railway passenger transport and is a best form to descrip volatility.For passenger,forecast and analysis of railway passenger flow can avoid delaying passengers' travel by helping passengers to arrange traveling plans and booking tickets time.For railway departments,forecast and analysis of railway passenger flow can provide the basis date of investment structure or operation and management or train dispatching decision,thus reasonable arrangement of trains and the number of train to avoid the waste of resources or passengers hard to get tickets.For our nation,it plays an important role in economic development pattern and the allocation of resources.So,under the background of the national transportation cause big development and the huge economic and social value it made,many scientific research workers and researchers pay more and more attention to for railway passenger volume prediction.In railway passenger traffic forecasting and analysis,time series model has been widely used,and also the current mainstream of forecasting methods.However,the existing railway traffic time series prediction method is often based on a single model or a simple extension of the model,although have compound model of two kinds,it is a simple mixture without considering the characteristics of the data and produce "deviation model design".In the process of railway passenger volume forecast cannot capture various features of the passenger flow data,leading to inaccurate results.So in order to meet the requirements of railway passenger flow forecast accuracy and fully grasping characteristics of railway passenger flow data.This paper proposed a new hybrid prediction analysis method based on the analysis of railway passenger traffic data.The main research work is as follows.First of all,according to the characteristics of the railway traffic data to analyze and research,Using Autoregressive Moving Average Model,Extreme Learning Machine Model,neural network Model and the seasonal Model respectively to analysis the character of railway passenger traffic data.They are widely used to linear,nonlinear and seasonal data characteristic research in the traditional time series models prediction.This paper adopted autoregressive moving average model,extreme learning machine,neural network and the seasonal model for railway passenger volume forecast and analyzed.After analysing the result of the experiment shows that: it is difficult for a single model to accurate predict the traffic time series data which is strong randomness,complex and unstable.By building hybrid algorithm model to predict railway passenger volume,we can see the result is improved,what's more,it is the current development trend of the time series prediction.The second,it's easily affected by many factors for railway traffic time series,it's strong randomicity,complex unstable.At the same time,we can't ignore the influence of the holidays for railway passenger volume.In order to better capture the railway passenger volume,give full play to the respectively advantages of nonlinear and liear models,we are going to regard railway traffic time series as a composite of high frequency and low frequency time series.The wavelet analysis theory and technology can decompose this kind of compound data into high frequency and low frequency components.Thus,this paper puts forward a railway passenger traffic prediction model base on wavelet decomposition of the extreme learning machine model and autoregressive moving average model,the railway traffic time series is decomposed into high frequency random component and trend of low frequency component by using the wavelet theory,then using the ELM model to forecast the high-frequency sequence to obtain the nonlinear railway passenger volume character and using the ARIMA model to forecast the low-frequency sequence to obtain the linear railway passenger volume character,at last use the high frequency sequence forecast predicted value and the low frequency predicted value to synthesis for railway passenger volume forecast results.In the model,the nonlinear high frequency random sequence is predicted by ELM model,the trend of low frequency linear sequence is predicted by the ARIMA model,the advantages of each model into full play,therefore the model for railway traffic time series prediction effect should be improved.Using the proposed hybrid model for railway passenger volume monthly data forecast show that,hybrid model predicted results average relative percentage error and root error are reduced than the results of the single forecasting model.It is improved on the prediction accuracy comparing with BP neural network model which is used to contrast in WeiWeiWang's paper of Traffic prediction research under the influence of high speed railway,and it is also some improved on the prediction precision compare with jiayuma who recent give complex seasonal time series model.The last,in the construction of railway passenger traffic mixed forecasting model and validation experiments,we further study found that,the railway traffic time series data has obvious seasonal characteristics,the hybrid model without considering the seasonal characteristics of railway passenger traffic,haven't reached the best result and there are further room for improvement.Based on the above situation,further expansion of the research based on the wavelet decomposition of extreme learning machine and autoregressive moving average mixed model,using the entropy method to deal with the forecast results of two kinds,and get the weight of these two kinds of forecast results,finally use the weighted average method to get the eventually railway traffic forecasting results.The experimental verification show that,the model of much feature fusion prediction model of railway passenger volume based on wavelet transform can further improve the prediction accuracy.From what has been discussed above,in this paper,the rail traffic time series hybrid prediction model and method not only has certain theoretical significance,but also has potential application prospects and market economic value in promoting and improving the complex time series prediction research.