Using Different Term Weighting Schemes of Centroid-Based Classifiers to Classify Drug Monographs

Abstract

With an increasing number of documents for drug monographs on the Internet, automatic classification of documents is an important task for organizing these documents into appropriated classes. The monographs of drug can be regularly categorized by their indications. A centroid-based classifier is a relatively high performance classifier with relatively less computation. To enhance the efficiency of standard centroid-based classifier with TFIDF to classify drug monographs, different term weighting schemes of a centroid-based classifier are evaluated. Moreover, the combination of a set of centroid-based classifiers with different term weighting schemes is proposed in this work. To evaluate the proposed method, two set of drug monographs are drawn from DailyMed and RxList websites are used. From the experimental results, the proposed method can improve the performance of the centroid-based classifier.