摘要
在总结前人工作基础上 ,提出了一种关于多维时间序列门限回归模型的通用建模新方案 ,并从理论上证明了此新表达式与经典门限回归模型的一致性。将此模型应用于云南五月雨量的预报中 ,其平均历史拟合和外推预报准确率分别为 71%和 6 8%。表明此通用建模方案可行的 ;
After summarizing the correspondence studies,the new method of building the threshold regression model of multidimensional time series has been advanced.The coherence between the new model and classical model has been proved theoretically.The new model was applied to forecast the May rainfall of Yunnan.The fitting and the forecasting accuracy of the new model are 71% and 68% respectively.The results indicated that the new method of building the threshold regression model of multidimensional time series is feasible and the new model has its applied value.
出处
《气象科学》
CSCD
北大核心
2002年第2期205-209,共5页
Journal of the Meteorological Sciences
基金
国家自然科学研究基金 ( D4 996 5 0 0 1 )的资助
关键词
非线性
多维时间序列
半截多项式
气候变化
Nonlinear Multidimensional time series Step polynomial Climatic change
