Research On Prediction Model Of Chaotic Time Series

In this paper, on the basis of analyzing the research status at home and abroad,With the logistic equation, simulation pest population chaotic time series and chaotictime series prediction center focal point, complete the following research and relatedprogramming:(1) Analyzed the agricultural neighborhood Logistic equation model and itschaotic time sequence generation parameters: initial value of0.22(between0and1), μvalue of4(between3and4), the length of the time series, such as N, is400. In thisstudy based on chaos theory, by definition explore the chaotic nature of chaotic timeseries Discriminant two angles: quantitative analysis and quantitative calculations.Which the law of quantitative analysis using the power spectrum method, sensitivedependence on initial conditions analysis; draw given time sequence power spectrumshows the time series of peaks together into one, to determine chaotic; quantitativecalculation rule correlation dimension, largest Lyapunov exponent calculation(WolfLaw and a small amount of data method). The methods through MATLAB programmi-ng were calculated associated dimension, such as D2, it’s1.1111and non-integer;Wolf method to calculate the largest Lyapunov exponent, such as λ1，it’s0.76503andgreater than0,with the help of λ1, judgment given parameter logistic time series withchaotic behavior, And the small amount of data to calculate the maximum Lyapunovexponent such as λ1，it’s0.65948and greater than0, get a same conclusion. Incomparison, the quantitative calculations obtained criterion of more convincing, andthen reinforce the basis of the data source for the prediction of chaotic time series.(2) Researched the selection of the phase space reconstruction parameters.Embedding dimension selection, auxiliary GP algorithm, a rough calculation embeddi-ng dimension, such as m, it’s3; The delay time is selected, focus through C-Cmethod. The method by calculating the time window to calculate the embeddingdimension and delay time, the results embedding dimension, such as m, it’s3,and theresults delay time, such as τ，it’s3.According to forecast data is calculated embeddingdimension and delay time, on Logistic chaotic time series phase space reconstruction,to deconstruct basic data for the prediction of chaotic time series, deconstruction3*394.(3) Construction of a chaotic time series prediction model based on neuralnetwork. Embedding dimension of the model by using the C-C,such as m, it’s3, As alink, to communicate the close relationship of the chaotic time series and neural netwo -rk.This model, the hidden layer nodes through dynamic comparison, selected theoptimal number of nodes7. Construct3-7-1of the topological structure of thenetwork;280training data set of100test data taken genetic algorithm to optimize theinitial rights of the BP neural network weight and threshold, the conclusion proved bythe genetic algorithm to optimize BP neural network, not only effectively improve thenetwork prediction accuracy, that’s GA-BP:1.1315e-07/mse, and single BP network,that’s BP:3.116e-06/mse, And to some extent, to strengthen the BP neural networkconvergence, that’s GA-BP:epoch(13), but single BP network, that’s BP:epoch(23).(4) In the MATLAB environment, has developed a model system based on thetheory of chaotic time series prediction. The system covers three subsystems, namely,the logistic equation model, chaos theory model and genetic BP algorithm model.