Abstract: In the midst of the mounting accessibility of electronic documents and the fast development of the World Wide Web turned out to be the enormous task of automatic categorization of documents. It has been developed into the key method for systematizing the discovery of information and knowledge. The appropriate categorization of text mining is required for the digital libraries, e-documents, blogs, emails and online news, machine knowledge and usual language processing methods to get a significant knowledge. The primary objective of this paper is to accentuate the significant methods and procedures that are engaged in manuscript documents categorization, at the same time creating awareness about some of the intriguing challenges that continue to be solved, chiefly focused on text representation and machine learning techniques In this study, we initiate a new-fangled technique which associates the characteristic selection and categorization techniques to pace up the text classification process and then about the low time consumption. In this paper, we propose a new method of IG-ANN which is the combination of feature selection and classification technique that increases and improves the classification accuracy, feature selection rate. We demonstrate the effective of our process by means of a systematic assessment and similarity over 13 datasets. The performance can be improved thus achieved makes ANN comparable or higher to supplementary classifiers. The projected algorithm is revealed to do better than the further conventional techniques like Best First Search wrapper method and filtered attribute method.