Prediction Methods Of Chaotic Time Series Based On SVM

Chaos time series prediction is the important research field and hot spots in today’ssociety. Chaos time series prediction is widely used in natural science and social science. Inthe electronic against, hydrological forecasting, image processing, ice age forecast, sunspotsand stock prices, etc of prediction. Chaos time series prediction has very important practicalapplication value and significance. The researches of chaotic time series prediction methodare many, but there is also research imperfect. This paper put forward in recent years theselection algorithm of support vector machine to the chaotic time series forecast. Thealgorithm of support vector machine used in classification, was just recently scholars into theprediction field, what the forecast effect is not a unified paper choose support vectormachine forecasting algorithm construct chaotic time series forecasting model. Use theprediction model of the three kinds of typical chaotic time series prediction research, andcompared it to the classical BP neural network predictive model and RBF neural networkmodel. The results of the study show that accuracy of use SVM algorithm predicts the Lorenzchaotic time series is the highest. And in prediction of the Henon chaotic time series andLogistic chaotic time series, RBF show effect is the best. This shows that SVM in predictionis not suitable for all the chaos time series, but in chaos time series Lorenz can get good paperis arranged as follow:In the first chapter, summarizes the background and development of the chaotic timeseries, Introduced the research status of the support vector machine (SVM) and the chaotictime series in the research progress. The second chapter, this paper introduces the theory ofchaos time series based and phase space technology. The third chapter, set up support vectormachine (SVM) model, choose the right parameter C and g. In the forth chapter, forecast thethree typical chaotic time series in the Matlab analyze the results. In the fifth Chapter, set upthe BP and RBF neural network forecast the three typical chaotic time series and analysis theresults. Compare of the result with the SVM results predicted.Through the analysis and comparison, we found in the Lorenz chaotic time seriesprediction, using SVM algorithm prediction accuracy is the highest. And in prediction ofHenon and Logistic chaotic time series the effect of RBF algorithm performance is the best.