加权马尔科夫AR-GARCH-GED模型在降水量中的预测  

The Application of Weighted Markov AR-GARCH-GED Model in the Prediction of Precipitation

作　　者：茹正亮[1] 杨芝艳[1] 朱文刚[1] 杨红莉[1] 

机构地区：[1]南京工程学院基础部,江苏南京211167

出　　处：《系统工程》2013年第12期98-102,共5页Systems Engineering

基　　金：南京工程学院校级青年基金资助项目(QKJB2011022);引进人才科研启动基金资助项目(YKJ201114)

摘　　要：针对降水量序列的随机性,基于非线性时间序列理论和马尔科夫链原理,首先应用AR-GARCH-GED模型拟合时序的总体趋势,得到的精度指标是随机波动的,其次应用有序聚类对精度指标分类,用加权马尔科夫链对精度指标状态预测,再次应用状态概率线性插值法修正精度指标,最后反推出预测值。该方法既包含了非线性时间序列的点预测,又融合了加权马尔科夫模型的状态预测,从而较好的揭示了降水量的内在规律,提高了预测精度,得到了满意的结果。Directed at the randomness of precipitation sequence based on nonlinear time series theory and Markov chain theory, the AR-GARCH-GED model is firstly applied, which fits the timing of the overall trend, and the resulting accuracy indicators with random fluctuations are obtained. Then sequential cluster is applied to classify the precision indexes, whose state is predicted by Markov chains. In addition, state linear interpolation is used to correct the accuracy index. Finally the predictive value is anti-inferred. The method including the point prediction of the nonlinear time series and the state prediction of weighted Markov model, which can further reveal the inner laws of precipitation and improve the accuracy of prediction as well as obtain the satisfactory results.

关 键 词：随机过程 加权马尔科夫 有序聚类 AR—GARCH—GED模型 预测 

分 类 号：O211[理学—概率论与数理统计]