Categorization of News Articles: A Model Based on Discriminative Term Extraction Method

Categorization techniques have major contribution in building automated system capable to fulfill the needs of decision making tasks for better organization and management of resources. The objective of this research is to assess the relative performance of some well-known classification methods. Among the proposed approaches our discriminative term extraction (DTE) based combined naive bayes and K-NN (NB-KNN) approach has the advantages of short learning time due to its computational efficiency with comparatively high accuracy. We designed DTE based NBKNN model for multi-class, single label text categorization. Our experiments suggest that data characteristics have considerable impact on the performance of classification methods. The Results obtained from Reuters-21578 corpus shows that NB-KNN consistently outperforms the single naive bayes and K-NN classifiers on Precision, Recall and Fl scores. The results of the study suggest designing a classification system in which several classification methods can be combined to increase the reliability, consistency and accuracy of the categorization.