Cascaded Class Reduction for Time-Efficient Multi-Class Classification

Abstract

Classification of objects into a fixed number of predefined categories has been extensively studied in wide variety of applications such as, for example, text categorization, web page classification, classi- fication of biological sequences, image recognition, speech recognition, and mining of business data, to name a few. These applications often involve hundreds or thousands of classes. Experiments show that approaches such as SVM and KNN often outperform other approaches, but suffer long classification time especially when the number of classes involved is large. In this paper, we investigate and propose a cascaded class reduction method in which a sequence of classifiers are cascaded to successively reducing the set of possible classes. We show that by cascading SVM and KNN with time-efficient classifiers such as linear classifiers and naive Bayes, we can significantly reduce the classification time of SVM and KNN while maintaining and sometimes improving their classification accuracy.