基于贝叶斯网络的不确定环境下多属性决策方法

A Bayesian-Networks-Based Method for Multiple Attributes Decision-Making under Uncertainty

融合贝叶斯网络推理技术来求解不确定多属性决策问题,根据问题的决策变量、环境变量和多个属性之间的依赖关系构造贝叶斯网络,通过推理求解在各个方案下每个属性取值的概率分布,从而把问题转化成风险决策问题.采用此方法求解不确定环境下多属性决策问题时,决策者只需考虑节点与其父节点之间的依赖关系,降低了思考的复杂程度,适用于大规模的复杂问题求解.算例表明基于贝叶斯网络的求解方法对不确定环境下的多属性决策问题是有效的.

A Bayesian-Networks-based method for multi-attributes decision making under uncertainty environment is proposed in this article. We construct Bayesian network based on the causality among decision variants, environment variants, and attributes. By the reasoning through the Bayesian network, the distribution of each attribute under each alternative can be calculated. Thus one complicated decision problem is modeled into a risky decision problem. By then, the decision makers can separately consider the relation between each node and its parents, and it＇s much easier than considering the distribution of attributes under the mass interactive influence factors condition. This method is appropriate for large scale, complicated problem. In the article, one sample is also proposed to demonstrate the application of this method.