Utilizing Category Relevancy Factor for Text Categorization

摘要

One of the main preprocessing steps for having a high performance text classifier is feature weighting. Commonly used feature weighting methods such as TF and IDF-based methods only consider the distribution of a feature in the documents) and do not consider class information for feature weighting. In this paper, we present TFCRF (Term Frequency and Category Relevancy Factor) method in which the weight of features depends on their power to discriminate the classes from each other by using class information. The results show significant improvement in the performance of SVM algorithm by using TFCRF feature weighting method in comparison to the other implemented standard feature weighting methods.