基于权重的朴素贝叶斯分类器设计与实现  

Design and Fabrication of Weight-based Nave Bayes Classifier

作　　者：代磊[1] 马卫东[1] 王凌楠[1] 马建国[1] 

机构地区：[1]西南科技大学信息学院,四川绵阳621010

出　　处：《情报理论与实践》2008年第3期440-442,456,共4页Information Studies:Theory & Application

基　　金：国家自然科学基金(项目编号:60272014);国家“863”计划项目(项目编号:2005AA121520);西南科技大学重点科研基金项目(项目编号:06zx2130)研究成果之一

摘　　要：贝叶斯分类器可以归结为求词条的先验概率,目前分类器中普遍使用词条的文档出现次数和词频来计算先验概率。本文提出了一种基于权重的朴素贝叶斯分类器,不仅改进了文本中词条的先验概率计算方式,并增加了词条的权重对计算的影响。该分类器使用TFIDF模型及其改进算法实现了分类器的设计。实验结果表明,该分类器的效果比传统算法有较大的改进。Bayes classifier can be summed up as finding the prior probability of words. At present, classifiers generally calculate the prior probability of words according to the times and the frequency by which the words appear in the documents. This paper introduces a weight - based Naive Bayes classifier, which can not only improve the - calculation of the prior probability of words in the documents, but also increase the impact of the weight of words on calculation. The classifier uses the TFIDF model and its improved algorithm to fulfill the design of the classifier. The results of the experiment indicate that this classifier has much more improvement than that by the use of the traditional algorithm.

关 键 词：文本分类 朴素贝叶斯分类器 算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]