Chaotic Time Series Analysis And The Research Of Its Application In Railway Freight Volume Prediction

With the development of nonlinear chaos dynamics, people have a better understanding ofthe complexity of time series, especially the chaotic time series analysis has become a veryimportant research direction. Description and prediction of railway freight volume time series isalways a very important research subject of railway freight volume system and it is an importantbasis of organizing railway development planning and optimizing resource allocation. It is alsoan important basis for economic evaluation and statistical analysis of railway constructionproject. Basing on the analysis of chaotic time series, chaos identification and prediction ofrailway freight volume time series are researched, the main results are as follows:(1) Basing on railway freight volume historical data and statistical analysis relevant theory,analyze the railway freight volume statistical method and statistical indexes. Use the RescaledRange Analysis method for analysis of self similarity, fractal dimension, long range correlationof railway freight volume time series.(2) On the basis of Takes theory, discusses the technology of phase space reconstruction oftime series, also compare several current estimation methods of time delay and embeddeddimension, and use these methods to conduct phase space reconstruction on the Lorenz equationand the railway freight volume time series. The reconstructed attractor of Lorenz equation ofrailway freight volume time series are obtained.(3) Identify the chaotic characteristics of the railway freight volume time series. Chaosidentification is conducted from the main component, saturation correlation dimension,maximum Lyapunov exponent and Kolmogorov entropy etc. Chaotic characteristics are alsoanalyzed. Basing on the different chaotic characteristics vectors, get the conclusion that therailway freight volume growth amount and growth rate time series are chaotic.(4) Because of the chaotic characteristics of railway freight volume growth amount andgrowth rate time series, conduct prediction of the railway freight volume basing on RBF neuralnetwork and BP neural network. And then the result is verified and analyzed.(5) On the basis of the analysis of the Volterra self-adaptive algorithm, analyze the impact on prediction accuracy of the number of Volterra self-adaptive model training data and thecharacteristic of using less training data to predict. Conduct Volterra prediction on railwayfreight volume. The error is verified and a series of meaningful results are obtained.The main innovation of this paper is:(1) The chaotic time series research method is introduced into the production of railwaytransportation science. The chaos theory is applied for analysis and research on the railwayfreight volume and a series of results are obtained.(2) Do research on the railway freight volume time series using RBF neural network thefirst time. The research steps of railway freight volume time series basing on RBF neuralnetwork are proposed and the result is compared with BP neural network prediction. And thenthe volume data of the next6years is obtained. Also the data is compared with the data got byBP neural network. The result shows that, the prediction data basing on RBF neural networkmatches the actual data better. The prediction accuracy of RBF neural network is much higherthe BP neural network. RBF neural network can be widely used in railway freight volume timeseries prediction.(3) Propose railway freight volume Volterra self-adaptive chaotic time series model, whichuses only one-dimensional time-series to get higher prediction accuracy, reducing therequirement of training data and improving the prediction efficiency of the algorithm. Itprovides a new approach in how to improve prediction accuracy in the case of limit railwayfreight volume historical data.(4) The models basing on RBF neural network and Volterra self-adaptive prediction enrichthe railway freight volume prediction methods. The result can be used as reference in railwayfreight volume trend prediction and railway planning.