基于节点块序列约束的局部贝叶斯网络结构搜索算法

针对K2算法过度依赖节点序和节点序搜索算法评价节点序效率较低的问题, 提出一种基于节点块序列约束的局部贝叶斯网络结构搜索算法, 该算法首先通过评分定向构建定向支撑树结构, 在此基础上构建节点块序列, 然后利用节点块序列确定每个节点的潜在父节点集, 通过搜索每个节点的父节点集构建网络结构, 最后对该结构进行非法结构修正得到最优贝叶斯网络结构.利用标准网络将算法与几种不同类型的改进算法进行对比分析, 验证该算法的有效性.

Local Bayesian Network Structure Searching Using Constraint of Node Chunk Sequence

The performance of the K2 algorithm depends on node ordering heavily. The Bayesian learning algorithm based on node order search needs to evaluate the quality of the node sequence by K2 algorithm, resulting in the problem of low efficiency in large and medium-sized networks. In this paper, a new Bayesian structure learning algorithm is proposed to solve the BN structure learning problem by searching local Bayesian network structure construction using node chunk sequence constraints. The algorithm firstly constructs a directional maximum weight spanning tree structure by using score orientation and generates the node chunk sequence by this structure. Then, the node chunk sequence is used to determine the potential parent node set of each node. The network structure is built by searching the parent node set of each node, and the structure is modified by illegal structure to get the optimal Bayesian network structure. Finally, some experiments are designed to evaluate the performance of the proposed algorithm, in which the standard network is used to compare the algorithm with several different improved algorithms to verify the effectiveness of the algorithm. The results indicate that the proposed algorithm is worthy of being studied in the field of BNs construction.