用贝叶斯网络进行因果分析

Bayesian Causal Analysis

因果分析是贝叶斯网络的一个重要应用领域。因果分析不同于相关分析,无论对数据分析、扰动分析还是预测都是十分重要的。贝叶斯网络虽然有一定的因果语义(我们常常用变量的因果关系构造贝叶斯网络结构),但贝叶斯网络是条件独立性的表示,因此我们不能不加限定地用贝叶斯网络进行因果分析。

The Bayesian causal analysis includes two techniques, one of which takes advantage of Bayesian network structure learning under the Causal Markov assumption and the presupposition that hidden variables are absent, and the other uses canonical form influence diagram. The two techniques possess their distinctive characteristics,and ought to be selected and put to use in the light of specific conditions.