基于有序样品聚类的集对权马尔可夫链年降水量预测模型

Set pair weight Markov chain model based on sequence clustering method for dynamically predicting annual precipitation

本文尝试将有序样品聚类、集对分析和马尔可夫链三种方法相结合,对传统的加权马尔可夫链预测方法进行了多方面改进,建立了基于有序样品聚类的集对权马尔可夫链年降水量预测模型,并将其应用于吉林省白城地区白城站2008-2010年年降水量的预测.将预测结果与实测值进行对比分析,可以发现:与传统方法相比,改进方法可以使降水量等级区间的划分更加合理,并增加了预测概率的集中程度,有效提高了预测精度.实测值均位于预测区间内,表明该方法具有较高实际应用价值.作为降水量预测模型的改进尝试,其预测效果令人满意.

A set pair weight Markov chain model based on sequence clustering method for predicting annual precipitation was established in this paper and applied to forecast the precipitation of Baicheng station （Jilin Province） during 2008 2010. It was an improvement of the traditional method by combining sequence clustering method, set pair analysis and Markov chain. Research results show that the improved method make the partition of precipitation grade interval more reasonable. In addition, it can effectively improve the concentration of prediction probability and the prediction accuracy. The measured values all lie in the prcdiction interval. In conclusion, the method is with high practical application value. As an attempt of the improvement of precipitation prediction model, its prediction effect is satisfactory.