基于遗传算法联姻策略的贝叶斯网络结构学习OA北大核心CSTPCD

The study of Bayesian network structure based on evolution methods suffers from local optimality and low efficiency.To address this issure,a Bayesian network structure learning method based on genetic algorithm marriage strategy is proposed.First of all,the"same"marriage strategy is designed.That is,two groups use the same search strategy and evaluate models to complete the Bayesian network structure learning.Then we marry the best quality sub-generation individuals,and iterate the best quality sub-generation individuals.Because the sub-generation of the marriage retains another group of segments,it has a good guarantee for the diversity of genes in the population,and effectively avoids the defects caused by the reproduction of close relatives.In response to the marriage strategy of the same agent model failing to ensure the quality of network structure and learning efficiency at the same time,we propose integrated genetic algorithm marriage strategies.Specifically,two groups use different agency models and search strategies to learn,and then the best-quality individuals in each group are married and iterated.Experiments show that the learning accuracy and effectiveness of the proposed algorithm on all-scale networks are better than the comparative algorithm.