基于关联分析的改进朴素分类器构造
首发时间:2010-11-10
摘要:朴素贝叶斯分类器是一种简单高效的分类器,但是其所依赖的属性独立性假设在真实问题中往往并不能成立。为了放松属性独立性约束来提高朴素贝叶斯分类器的泛化能力,研究人员已经进行了大量的工作。本文提出了基于关联分析的改进朴素贝叶斯分类器构造方法(BCCA)。在训练阶段,BCCA找到所有的频繁闭项集。在测试阶段,BCCA对于测试样本所包含的每个频繁闭项集构造一个分类器,通过集成这些分类器来给出预测结果。实验验证了该算法的可行性与有效性。
For information in English, please click here
Na?ve Bayesian Classifier construct based on Correlation Analysis
Abstract:Naive Bayesian classifier is simple and efficient. However, it strictly relies on the assumption that all the attributes are independent. But this assumption is usually can't be met in the real problems. In order to improve the generalization ability of Naive Bayesian classifier by relaxing the independence constrains between attributes, researchers have done a lot of work. In this paper, an improved Naive Bayesian classifier constructor (BCCA) based on correlation analysis is proposed. In the training phase, BCCA find all the frequent closed itemsets. In the test phase, for each closed frequent itemset that is contained by the test sample, BCCA construct a classifier respectively. The predict result is given by integrating these classifiers. The experiments have verified the feasibility and effectiveness of BCCA.
Keywords: Naive Bayesian Correlation Analysis frequent itemset ensemble learning
论文图表:
引用
No.4390301550045128****
同行评议
共计0人参与
勘误表
基于关联分析的改进朴素分类器构造
评论
全部评论0/1000