基于马氏链拟合的一种非负变权组合预测算法及其应用  

An Algorithm and Its Application for Fitting Combinatorial Forecasting with Nonnegative Time-varying Weights Based on Markov Chain

作　　者：张新育[1] 李镇[1] 周世国[1] 

机构地区：[1]郑州大学数学系,河南郑州450052

出　　处：《大学数学》2009年第1期70-74,共5页College Mathematics

基　　金：国家自然科学基金资助项目(10671183)

摘　　要：通过马氏链拟合的方法求取一种新的非负时变权组合预测算法公式.主要工作是:一、对组合预测问题以最小误差为准则给出了马氏链的状态和状态概率初步估计;二、用马氏链拟合状态概率分布时变规律,通过约束多元自回归模型导出了一步转移概率阵的LS解;三、给出一种非负时变权组合预测公式并举一应用实例.Our purpose in this paper is to development a new algorithm formula to fit combinatorial forecasting with nonnegative time-varying weight by use of Markov chain. The main results is as below： 1. Deriving initial status and its probability estimation of Markov chain under least squares criterion according to combinatorial forecasting problem; 2. Fitting the law of status probabilities distribution varying with time by use of Markov chain, then deriving the least squares solution to one-step status probabilities transition matrix by means of constrained multivariate self-regression model 3. Deriving a formula for combinatorial forecasting estimation with nonnegative time-varying weights and make an living example in application.

关 键 词：组合预测 非负时变权 马氏链 约束多元自回归模型 

分 类 号：F204[经济管理—国民经济] F224.0