用于风险管理的贝叶斯网络学习

结合专家知识和数据进行贝叶斯网络学习.首先利用专家知识建立初始贝叶斯网络结构和参数;然后基于变量之间基本依赖关系、基本结构和依赖分析方法,对初始贝叶斯网络结构进行修正和调整,得到新的贝叶斯网络结构;最后将由专家和数据确定的参数合成为新的参数,得到融合专家知识和数据的贝叶斯网络.该方法可避免现有的贝叶斯网络学习过于依赖数据、对数据的数量和质量要求过高等问题.

Learning Bayesian networks in risk management

A new method of learning Bayesian networks is presented, which can effectively combine expert knowledge and data. Firstly, an initial Bayesian network structure is set up by using expert knowledge. Then, it is revised and regulated based on basic dependency relationship between variables, basic structure between nodes and dependency analysis method to obtain a new Bayesian network structure. Finally, two kinds of parameters got respectively by expert knowledge and data are fused to produce new parameters, and a Bayesian network combining expert knowledge and data is gained. This method can avoid the problems of depending on a large number of data with high quality in existing Bayesian network learning.