The traditional time series prediction algorithm can well simulate and predict the stable time series data, but not so well to the series of nonlinear and non-stationary. To solve this problem, an improved algorithm comes up. Through the wavelet decomposition and single reconstruction, the original time series is decomposed into a layer of low frequency data and two layers of high frequency data. The GARCH model is used to forcast the low frequency data, the improved algorithm is used to forecast the two layers of high frequency data. Through Markov model predicting the state interval, with the smoothing coefficient, the high frequency data is predicted. The final forecasting result comes from the superposition of the three layers of prediction result. Through the error test, the accuracy of the improved algorithm has a major improvement.%时间序列的传统预测方法能够很好地拟合和预测平稳时间序列，对于非线性非平稳的时间序列数据预测效果不好。为解决该问题，文本提出一种改进的预测算法。通过小波分解和单边重构，原始时间序列被分解为一列低频数据和两列高频数据。低频数据采用传统的时间序列方法GARCH模型预测，高频数据使用改进方法预测。通过马尔科夫模型预测出状态区间，结合指数平滑法，预测出高频结果。与低频数据结果叠加得到最终预测结果。经误差比较，改进算法预测精度有较大提升。
展开▼
