文章摘要
基于微生物遗传算法的贝叶斯网络结构学习
Bayesian Network Structure Learning Based on Microbial Genetic Algorithm
投稿时间:2020-09-25  修订日期:2020-09-25
DOI:
中文关键词: 贝叶斯网络  结构学习  最大信息系数  微生物遗传算法
英文关键词: bayesian network  structure learning  maximum information coefficient  microbial genetic algorithm
基金项目:国家自然科学基金项目(71762028);新疆维吾尔自治区社会科学基金 (19BTJ036);新疆财经大学科研创新项目 (XJUFE2020K007)。
作者单位E-mail
武梦娇 新疆财经大学统计与数据科学学院 708875911@qq.com 
刘继 新疆财经大学统计与数据科学学院 liuji5000@126.com 
摘要点击次数: 100
全文下载次数: 
中文摘要:
      网络评论信息表达随着人工智能的发展,贝叶斯网络受到广泛关注。结构学习作为贝叶斯网络学习的重要内容,存在学习效率低、容易陷入局部最优等问题。本文提出了一种贝叶斯网络结构学习改进算法(MIC-MGA),首先利用最大信息系数得到初始种群,然后利用微生物遗传算法中的选择、交叉和变异操作算子优化初始种群。在学习过程中将贝叶斯信息准则作为适应度函数,通过学习训练得到最终的贝叶斯网络结构。实验结果表明,本文算法在小样本数据情况下能学习到更接近真实网络的网络结构,具有较好的学习性能。
英文摘要:
      With the development of artificial intelligence, Bayesian networks are widely concerned. As an important part of Bayesian network learning, structure learning has many problems, such as low learning efficiency, easy to fall into local optimum and so on. In this paper, an improved Bayesian network structure learning algorithm (MIC-MGA) is proposed. First, the initial population is obtained by using the maximum information coefficient, and then the initial population is optimized by using the selection, crossover and mutation operators in the microbial genetic algorithm. In the learning process, the Bayesian information criterion is used as the fitness function, and the final Bayesian network structure is obtained by learning and training. The experimental results show that the algorithm can learn the network structure closer to the real network in the case of small sample data, and has better learning performance.
View Fulltext   查看/发表评论  下载PDF阅读器
关闭

手机扫一扫看 分享按钮