A Novel Approach in Feature Selection Method for Text Document Classification

In this paper, a novel approach is proposed for extract eminence features for classifier. Instead of traditional feature selection techniques used for text document classification. We introduce a new model based on probability and over all class frequency of term. We applied this new technique to extract features from training text documents to generate training set for machine learning. Using these machine learning training set to automatic classify documents into corresponding class labels and improve the classification accuracy. The results on these proposed feature selection method illustrates that the proposed method performs much better than traditional methods.