Prediction comparative study of complex multivariate systems with AGA-BP

To improve the prediction accuracy of complex nonlinear systems(such as chaotic systems, power load and stock market), a novel scheme formed on the basis of AGA-BP neural network is proposed. According to Takens Theorem, nonlinear chaotic time series is reconstructed into vector data, AGA-BP neural network is used to fit the trained data of the predicted complex chaotic system, then the network parameters of data matrix built with the embedding dimensions are estimated, and the prediction value is also calculated. To evaluate the results, the proposed multivariate predictor based on AGA-BP neural network is compared with univariate one with the same numerical data. The simulation results obtained by the Lorenz system show that the prediction mean squared error of the multivariate predictor is much smaller than the univariate one.