Empirical mode neural network-based chaotic time series prediction method

Abstract

The invention discloses a chaotic time series prediction method based on an empirical mode neural network, which belongs to the field of chaotic time series prediction. It includes the following steps: 1) Discrimination of chaotic characteristics of chaotic time series; 2) Preprocessing of chaotic time series data; 3) Reconstruction of phase space of chaotic time series; 4) Construction of empirical modal neural network and training of empirical modal neural network model; 5) Use the empirical mode neural network to predict the chaotic time series. Compared with the traditional chaotic time series prediction method, this method enhances the adaptive ability of the prediction model, and can adaptively construct the hidden layer activation function of the empirical mode neural network according to the characteristics of the data set itself, without any selection of activation function At the same time, the number of components of the sequence undergoing empirical mode decomposition is used as the number of hidden layers of the empirical mode neural network, which provides a new solution for the selection of the number of hidden layers.

Description

technical field [0001] The invention relates to the field of chaotic time series forecasting, in particular to a chaotic time series forecasting method. Background technique [0002] Chaotic time series forecasting generally exists in hydrology, economy, sunspot and stock market forecasting, but due to the complexity of chaotic time series and its initial value sensitivity, it is very difficult to accurately predict chaotic time series. [0003] The traditional prediction methods of chaotic time series include: global method, local method, adaptive prediction method and so on. Among them, the self-adaptive prediction method includes the prediction method based on neural network and the prediction method based on series expansion. The theoretical basis of these methods is the Takens embedding theorem, which obtains the relatively regular evolution trajectory of the chaotic system through phase space reconstruction, and then realizes the prediction of chaotic time series thro...

AI Technical Summary

Problems solved by technology

[0005] Nevertheless, the wavelet neural network still has its own limitations. For example, the selection of wavelet basis functions is crucial to wavelet decomposition, and how to choose a suitable basis function requires a certain experience of chaotic time series in advance, which leads to wavelet The adaptability of the neural network is limited, and there is a problem that the number of hidden layers of the neural network is difficult to determine

Images