【摘要】 针对单词簇上训练朴素贝叶斯文本分类器概率估计偏差较大所导致的分类精度较低问题,在概率分布聚类算法得到的单词簇的基础上,根据单词与簇间互信息建立有序单词子序列,采用有放回随机抽样对序列构造规模相当的样本集,并将估计出的参数的平均值作为训练得到的参数对未知文本进行分类。公共文本实验数据集上的实验结果表明,该文提出的训练方法相对于传统的朴素贝叶斯分类器训练方法能够获得更高的分类精度且过程相对简单。更多还原

【Abstract】 Aiming at the problem of low classification accuracy caused by poor distribution estimation for training naive Bayes text classifier on word clusters,a word list based on mutual information between word and clusters is made,a sample set with the same size with bootstrap method is constructed,and the average of the parameters estimated from it as the last parameter to classify unknown text is used.Experiment results on benchmark text dataset show that the method gain higher classification accuracy compared to naive Bayes classifier.更多还原