Prediction of time series based on wavelet decomposition and clustering fuzzy systems
基于小波分解和聚类模糊系统的时间序列预测

A prediction method for non-stationary time series was proposed in association with multi-resolution of wavelet analysis and interpretability of fuzzy rules. The original time series were decomposed into the smooth and the detailed at different levels. After being denoised with the method of soft - hard threshold value, the smooth and the detailed at different levels were forecasted with the clustering fuzzy systems. Finally the sum of the forecasting results at different levels was the prediction of the original time series. Experiments show that the method is effective.