An Integrated and Improved Approach to Terms Weighting in Text Classification

Traditional text classification methods utilize term frequency (ti) and inverse document frequency (idf) as the main method for information retrieval. Term weighting has been applied to achieve high performance in text classification. Although TFIDF is a popular method, it is not using class information. This paper provides an improved approach for supervised weighting in the TFIDF model. The tfidf-weighting model uses class information to compute weighting of the terms. Hie model also assumes that low frequency terms are important, high frequency tenus are unimportant, so it designs higher weights to the rare terms frequently. So, it uses rare term information along with class information for weighting. So, the paper proposes an improved approach which combines the benefits of the traditional kNN classifiers and Naive Bayes supervised learning method.