Efficient Categorization of Document with J48 Multi-Class Classifiers

In the text categorization field number of feature selection metrics has been explored such as Chi-Square (CHI), Information Gain (IG), Odds Ratios (OR), and Correlation Coefficient (CC). Form this OR and CC are the one-sided metrics and CHI and IG are the two-sided metrics. In this one side feature selection metrics only select the most indicative features and in the two side metrics only implicitly features are selected. Automated feature selection is important for text classification to decrease the feature size and to speed the learning performances of classifiers. In this paper present the feature selection approach for the efficient categorization of document based on the J48 algorithm. In proposed system we are working on a data set named as a 20-Newsgroups dataset which contains the data related to different news category this data set is taken as input to the system and as an output, we will get documents classification in several categories (topic) as an output. Here, the compared performance of the proposed system with existing Naive Bayes algorithm and the proposed system gives more accurate result as compared to the existing.