基于概率图的三支决策模型研究  

Three-way Decision Model Based on Probabilistic Graph

作　　者：薛占熬[1,2] 王朋函 刘杰[1,2] 朱泰隆[1,2] 薛天宇[1,2] 

机构地区：[1]河南师范大学计算机与信息工程学院,新乡453007 [2]"智慧商务与物联网技术"河南省工程实验室,新乡453007

出　　处：《计算机科学》2016年第1期30-34,共5页Computer Science

基　　金：国家自然科学基金计划项目(61273018);河南省基础与前沿技术研究计划项目(132300410174);河南省教育厅计划项目(14A520082);新乡市重点科技攻关计划项目(ZG14020)资助

摘　　要：概率图模型是一类用图形模式表达基于概率关系的模型的总称,用该模型解决损失代价问题已成为当前的研究热点。结合概率图和三支决策理论,提出了基于概率图的三支决策模型。该模型通过对数据进行分析,构造其Bayes网络;并根据模型中节点的相互依赖关系,计算出条件概率分布函数;结合查询变量的先验概率和三支决策损失代价函数,建立了相应的决策规则,给出了概率推理决策中代价最小化问题的一种解决方法。最后通过教学评估实例验证了该模型的有效性。Probability graph model is a general term for a class of graph patterns that express the probabilistic model,and the problem of solving the loss cost by using the model has become a hot research hotspot.In this paper,the threeway decision model was put forward based on probability graph and three-way decision theory.By analyzing the information data set,the Bayesian network of the model is constructed.Then according to the interdependencies of the nodes in the model,the conditional probability distribution function is calculated.And combining with the prior condition of the data set and three decision-making cost loss function,its decision rules are established.The method to solve the problem of cost minimization is given in probability reasoning decision-making.Finally,a teaching evaluation illustration was presented to verify the effectiveness of model.

关 键 词：三支决策 概率图模型 BAYES网络 精确推理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]