Text Classification Based on Naive Bayes with Adjusted Weights via Frequency Ratio of Feature Words
Zhaoyi Guo
Abstract
Text Classification has become a study hot spot because of its wide application in daily life. Many methods are proposed for this task, such as those based on various neural network models. However, the structure of neural network model is complex, so it is very troublesome to modify it. For text classification, there is one of the simplest and most classic methods, which is Naive Bayes. The purpose of this study is to adjust the weight of feature words to improve the effect of Naive Bayes, according to the different frequency ratio of feature words in the target category and other categories. Experimental results show that the performance of the modified Naive Bayes method is generally better than that of the original Naive Bayes method. Specifically, the performance of the two methods is similar without considering smoothing. In the case of considering smoothing, the new method is better than the old one.