基于因果影响独立模型及其中形成的特定上下文独立关系,提出一种适于样本学习的贝叶斯网络参数学习算法.该算法在对局部概率模型降维分解的基础上,通过单父节点条件下的子节点概率分布来合成局部结构的条件概率分布,参数定义复杂度较低且能较好地处理稀疏结构样本集.实验结果表明,该算法与标准最大似然估计算法相比,能充分利用样本信息,具有较好的学习精度. Based on the independent model of causal influence and the specific context-dependent relationship among them, this paper proposes a Bayesian network parameter learning algorithm which is suitable for sample learning.Based on the dimensionless decomposition of the partial probability model, , The conditional probability distribution of local structure is synthesized and the parameter definition complexity is low and the sparse structure sample set can be well processed.The experimental results show that this algorithm can make full use of the standard maximum likelihood estimation algorithm Sample information, with good learning accuracy.