从数据中学习贝叶斯网络往往会因为搜索空间庞大而耗费大量时间.由于贝叶斯网络固有的因果语义,领域专家往往能够凭借自己的经验确定节点之间的因果关系.本文方法充分收集专家的意见,并利用证据理论进行综合,去除无意义的网络结构,然后利用常用的学习算法从数据中继续学习.这种融合知识和数据的贝叶斯网络构造方法利用专家知识来缩小学习算法的搜索空间,避免了盲目搜索,同时也避免了单个专家知识的主观性.实验表明该方法能够有效提高学习效率. Learning Bayesian networks from data often consumes a lot of time because of the huge search space.Because of the inherent causal semantics of Bayesian networks, domain experts often can determine the causal relationship between nodes by their own experience.This method fully collects experts And use evidence theory to synthesize, remove meaningless network structure, and then continue to learn from the data using commonly used learning algorithms.This Bayesian network construction method that combines knowledge and data uses expert knowledge to reduce the learning algorithm Search space, avoid blind search, but also avoid the subjectivity of a single expert knowledge.Experiments show that the method can effectively improve the learning efficiency.