Sequential Classifiers Combination for text categorization: An experimental study

In this paper, we introduce Sequential Classifiers Combination (SCC) into text categorization to improve both the classification effectiveness and classification efficiency of the combined individual classifiers. We apply two classifiers sequentially for experimental study, where the first classifier (called filtering classifier) is used to generate candidate categories for the test document and the second classifier (called deciding classifier) is used to select a final category for the test document from the candidate categories. Experimental results indicate that when combining boosting and kNN methods, the combined classifier outperforms the best one of the two individual classifiers, and in the case of combining Rocchio and kNN methods, the combined classifier performs equally well as kNN while its efficiency is much better than kNN and is close to that of Rocchio.