混沌时间序列预测方法

针对混沌时间序列预测问题,深入分析了基于径向基神经网络与基于记忆库两种预测方法,进而应用两种方法分别对Logistic序列、Kent序列进行预测,仿真结果表明:输入节点个数严重影响RBF神经网络预测性能,样本容量的增大与预测精度提高只是在一定范围内呈正比;选取系统的关联维数作为预测阶数时,基于记忆库预测方法的平均预测误差较小,且其预测误差随样本容量的增多而减小.

Study on prediction technology of chaotic series

To solve the problem of the prediction of the chaotic series, we studied two prediction methods, one is based on RBF neural network and the other is memory - based predictor. These methods were used respeetively to the prediction of Logistic series and Kent series. The simulation results showed that the prediction capability of RBF neural network depends greatly on the number of the input nodes, and the increase of the sample size is in direct proportion to prediction precision only in a certain scope. When the predictive bank number is chosen as the correlation dimension of the system, the average prediction error of the memory based predictor is small, which also decreases as the sample number increases.