提出一种混合式贝叶斯网络结构增量学习算法.首先提出多项式时间的限制性学习技术,为每个变量建立候选父节点集合;然后,依据候选父节点集合,利用搜索技术对当前网络进行增量学习.该算法的复杂度显著低于目前最优的贝叶斯网络增量学习算法.理论与实验均表明,所处理的问题越复杂,该算法在计算复杂度方面的优势越明显. A hybrid Bayesian network structure incremental learning algorithm is proposed.Firstly, polynomial time constrained learning techniques are proposed to establish a set of candidate parent nodes for each variable. Then, based on the set of candidate parent nodes, the current network is made using search technology Incremental learning.The computational complexity of this algorithm is significantly lower than that of the current optimal Bayesian network incremental learning algorithm.Theoretical and experimental results show that the more complex the problem is, the more obvious the computational complexity of the algorithm is.