A Short-Term Prediction Method With BR-BP-JOS
Xueting Ma, Fang Liu, Tianzeng Zhao · 2025
The short-term forecasting of chaotic time sequences has the problem of low prediction accuracy. To improve the accuracy of short-term prediction of chaotic time sequences, a short-term prediction method combining Bayesian regularization double-layer back propagation network with joint order phase space reconstruction (BR-BP-JOS) is proposed. First, the sequence is preprocessed to make the statistical features of the sequence easier to be modeled and extracted. Then the sequence is reconstructed in phase space by joint order. Finally, a double-layer backpropagation network trained by Bayesian algorithm is used to predict the chaotic sequence. The suggested method is superior in short-term forecasting of chaotic time series, as shown by the experimental results and comparison analysis.