基于强化学习的贝叶斯网络模型生成方法研究
BAYESIAN NETWORK MODEL GENERATION METHOD BASED ON REINFORCEMENT LEARNING
岑岗 1郑泽宇 1岑跃峰 1王佳晨 1吴思凡1
作者信息
- 1. 浙江科技学院信息与电子工程学院 浙江 杭州 310023
- 折叠
摘要
传统贝叶斯网络的网络结构需要人为事先确定,用于预测时模型可靠性与准确性较低,因此提出一种基于强化学习的贝叶斯网络模型生成方法.将强化学习用于对最优泛化残差评分的搜寻,通过构建邻接矩阵的方式将贝叶斯网络抽象成有向无环图;对于构建完成的贝叶斯网络,提出一种基于因果方向判断的贝叶斯网络结构优化方法.实验结果表明,该方法优于各类传统的贝叶斯网络结构生成方法.
Abstract
The network structure of the traditional Bayesian network needs to be determined in advance,and the reliability and accuracy of the model are low when used for prediction.Therefore,a Bayesian network model generation method based on reinforcement learning is proposed.The reinforcement learning was used to search for the optimal generalization residual score,and the Bayesian network was abstracted into a directed acyclic graph by constructing an adjacency matrix.For the completed Bayesian network,a Bayesian network structure optimization method based on causal direction judgment was proposed.The experimental results show that the method in this paper is superior to all kinds of traditional Bayesian network structure generation methods.
关键词
贝叶斯网络/强化学习/并行集成/因果方向判断/结构生成与优化Key words
Bayesian network/Reinforcement learning/Parallel integration/Causal direction judgment/Structure generation and optimization引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据