Scholarlink

<p>Bayesian networks (BNs) are a graphical model that represents probabilistic dependencies between variables, providing an intuitive and effective framework for dealing with uncertain information. In order to solve the problem of low accuracy of hybrid BN structure learning algorithms, a hybrid BN learning algorithm based on approximate skeleton is proposed in this paper. The algorithm first learns an approximate global skeleton, firstly, when performing skeleton learning, it determines whether the learned parent–child nodes in the dataset are symmetric or not. Secondly, for the symmetric nodes, the AND rule is directly used; and for the non-symmetric nodes, the Markov blanket is used for checking them, and then the AND rule or the OR rule is used. And lastly, the hill-climbing algorithm is utilized for determining the direction of the directionless edges in the skeleton. Experiments are conducted under 10 BNs with different sample sizes, and the algorithm proposed in this paper improves the accuracy and efficiency of this algorithm compared with the results of the three existing algorithms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A algorithm based on approximate skeleton of Bayesian network structure learning

  • 期刊论文

  • 发表日期：

    2026年2月28日

查看全文
  • Haoran Liu,
  • Ye Gao,
  • Qianrui Shi,
  • Niantai Wang,
  • Lin Li

摘要

Bayesian networks (BNs) are a graphical model that represents probabilistic dependencies between variables, providing an intuitive and effective framework for dealing with uncertain information. In order to solve the problem of low accuracy of hybrid BN structure learning algorithms, a hybrid BN learning algorithm based on approximate skeleton is proposed in this paper. The algorithm first learns an approximate global skeleton, firstly, when performing skeleton learning, it determines whether the learned parent–child nodes in the dataset are symmetric or not. Secondly, for the symmetric nodes, the AND rule is directly used; and for the non-symmetric nodes, the Markov blanket is used for checking them, and then the AND rule or the OR rule is used. And lastly, the hill-climbing algorithm is utilized for determining the direction of the directionless edges in the skeleton. Experiments are conducted under 10 BNs with different sample sizes, and the algorithm proposed in this paper improves the accuracy and efficiency of this algorithm compared with the results of the three existing algorithms.