Improved Bayes Method Based on TF-IDF Feature and Grade Factor Feature for Chinese Information Classification
Zhaowei Qu, Xiaomin Song, Shuqiang Zheng, Xiaoru Wang, Xiaohui Song, LI Zu-quan · 2018
Existing methods improved the accuracy of Bayes by weakening its feature independence assumption. However, these approaches only simply incorporate the learned feature into the formula of Naive Bayes, but they do not incorporate these features into its conditional probability. In addition, these feature weighting methods have received less attention and whose accuracy for information extraction and Chinese text classification still needs to be improved. In this paper, we propose a more effective and more accurate method for automatic information classification, called improved Bayes method based on TF-IDF feature weight and grade factor feature weight (TIGFIB), which estimates the conditional probabilities of Naive Bayes by TFIDF feature and imports grade factor feature into formula of Naive Bayes. Besides, we apply our improved Bayes method to Chinese text classification. Experiment shows that our improved Bayes method is superior to other feature weighting Naive Bayes methods.