基于Rough Set的贝叶斯网络结构学习研究

RESEARCH ON BAYESIAN NETWORKS STRUCTURE LEARNING BASED ON ROUGH SET

Rough Set理论与方法是处理复杂系统的一种有效方法,但未能包含处理不精确或不确定原始数据的机制,与贝叶斯网络等不确定性理论有很强的互补性.本文提出基于Rough Set理论的贝叶斯结构学习方法,把Rough Set理论与贝叶斯网络相结合,通过属性约简简化贝叶斯网络结构变量,更好满足条件属性间的独立性限制,降低结构复杂度;同时,条件属性之间的依赖性决定贝叶斯网络变量之间的依赖关系和弧的方向.最后,通过算例说明该方法的应用过程.

Rough Set theory and method is one kind of effective method for dealing with complicated system, but it fails to contain the theory handling uncertainty problem such as imprecise or uncertain data mechanism. So, it has strong complementarity with Bayesian network theory. The paper puts forwards a knid of Bayesian networks structure learining method combining Rough Set theory with Bayesian networks. The method reduces Bayesian networks variables by attribute reduct, improving the complexity of Bayesian network structure and meeting several independence between condition attribute; At the same time, the dependence between condition attributes is to be used for deciding dependency relationship and the arc direction between Beiyesi network structure variable. Finally, It provides an example on Bayesian networks structure learining to explain the application procedure of the approach.