基于遗传算法的贝叶斯网络模型研究

Research of Bayesian network model based on genetic algorithm

目前贝叶斯网络缺乏支持结构建立、参数学习、知识推理的一致算法,使知识建立与应用过程无法联接。针对这一现状,通过设计适合于贝叶斯网络学习的遗传算法编码方式、具有调整策略的交叉与变异算子,能进行推理误差反馈的适应函数,实现样本支持下的结构确定、参数学习、推理检验、反馈修正的贝叶斯网络全过程建立。实验结果表明,新算法不仅同步优化网络结构与参数,且可以自适应推理误差的学习修正,有着更满意的知识推理正确率。

The research of Bayesian network （BN） is still lacked of algorithms which can implement the consistence of structure building, parameter learning and knowledge inference, so that knowledge construction and application are out of association and verification. A novel algorithm is proposed, which designs a new BN learning encodement, crossover and mutation operators with adjust strategies and the fitness funetion with inferential error feedback, implements BN building in all processes ofstructure building, parameterlearning, knowledge inference and feedback revise with samples supported. Results show that the new approach can not only optimize the structure and parameters learning synchronously, but also revise inferential error adaptively, and has more satisfying and accurate inference result.