层次朴素贝叶斯分类器构造算法及应用研究

Construction and application of hierarchical nave Bayesian classifier

贝叶斯网络分类器是数据挖掘与知识发现领域研究的主要方法之一。层次朴素贝叶斯分类器通过引入潜在节点来实现属性变量间存在聚集的层次关系,提出学习该分类器的构造算法。算法首先借助节点间的条件互信息值来锁定可能聚集节点的范围,然后再通过模拟退火算法来搜索评分较高的模型。层次朴素贝叶斯分类器的结构特点适于构造水质富营养化评价模型,应用于水质预警系统的结果证明该方法可行,并具有较好的适用效果。

Bayesian network classifier is one of the main research methods in data mining and KDD domain. Because hierarchical na＇fve Bayesian （HNB） classifier can reflect the relations between variables by introducing latent aggregating nodes, this paper proposes a learning algorithm for modeling the classifier. Firstly the conditional mutual information between variables is used to derermine the scale of latent aggregate nodes, then simulation anneal （SA） algorithm is used to search the classifier with higher score. The structure characteristics of HNB are suitable for building water quality eutrophication model to realize the assessment and early warning of water quality. The algorithm proposed in this paper has been applied in a water quality security early warning system. The results show the feasibility and applicability of the model.