A PMLP Based Method for Chaotic Time Series Prediction
| Authors: | Yang Hongying, Tsinghua University, China
Ye Hao, Tsinghua University, China
Wang Guizeng, Tsinghua University, China
Zhong Maiying, Tsinghua University, China |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Time Series Modelling |
|---|
| Keywords: | chaos theory, time-series analysis, phase space, prediction, neural nets, parallel networks |
|---|
Abstract
This paper proposes a new method for prediction of chaotic time series based on Parallel Multi-Layer Perceptron (PMLP) net and dynamics reconstruction technique. The PMLP contains a number of multi-layer perceptron (MLP) subnets connected in parallel. Each MLP subnet predicts the future data independently with a different embedding dimension. The PMLP determines the final predicted result according to the weighted average of all sub-outputs. Simulation results show the effectiveness of the method.
