数据库信息分类中贝叶斯网络模型的应用

数据库信息分类中 ,朴素贝叶斯分类模型是一种简单而有效的分类方法 ,但它的属性独立性假设使其无法表达属性变量间存在的依赖关系 ,影响了它的分类性能 .而一般贝叶斯网络模型则由于能表达属性变量之间的依赖关系而越来越受到人们的重视 ,但一般贝叶斯网络分类模型结构的学习算法是一个NP完全问题 .本研究在一种简化的贝叶斯网络分类模型的基础上 ,利用其多项式时间复杂度的结构学习算法 ,将其应用于数据库信息分类 ,实现了学习效率和分类精度的一种折衷 .实验结果表明 ,这种分类方法有着比较高的数据库信息文本检索性能 .

The application of Bayesian network model in database information classification

In database information classification, Naive Bayesian Classification Model is a simple but efficient solution. However, the hypothesis that its attributes should be independent prevents it from expressing the dependences among the attribute variables, which affects the efficiency of classification greatly. So common Bayesian network Model, which can express the dependencies among attribute variables, is more and more important. And yet, the learning algorithm of the structure of common Bayesian classification model is NP-hard. In this paper, based on a simplified Bayesian network classification model, we apply its structure learning algorithm, with polynomial time complexity, to the classification of database information, and get a compromise between the learning efficiency and classification precision. The experimental result shows that this classification method has better performance in text retrieval of database information.