Text classification using machine learning algorithms

Abstract

Financial, educational and communal activities produce enormous amount of data. Automatic text classification has significant application in content organization, point of view extraction, evaluation analysis, spam filtering and sentiment analysis. Automatic classification of text documents requires information extraction, machine learning and Natural Language processing. We have proposed a probabilistic framework for text classification. Proposed classification model is composed of three major modules i.e. pre-processing of unstructured text, learning of probabilistic model and the classification of unseen data by using learned model. This framework is trained and tested by using “20 newsgroup” dataset containing twenty different news categories i.e. politics, sports, religions and pc hardware. We have used both supervised and unsupervised algorithms to get the full insight on the relationships among various text classification techniques. Highest accuracy of 84.51% was achieved for 4 categories by Naïve Bayes among the other Supervised Algorithms we used and 62.79% homogeneity was achieved for unsupervised algorithms for 4 categories which demonstrates the effectiveness score of proposed automatic text classification approach.