Classification of text documents using adaptive robust classifier

Abstract

Classifying the documents by means of extracting the keywords has become an imperative direction of research in text mining. The important purpose of extracting the keywords is to exemplify the documents in a concise manner. The compactable exemplification of documents serves multiple applications in different ways. Classifying the documents regards to the keywords have becomes a major task. Most classifiers are suitable only for the dataset which hold the low number of documents. In this paper, adaptive robust classifier (ARC) is proposed to classify the documents in any size dataset with better accuracy. ARC is designed to segregate the documents dataset into multiple parts and perform classification in a random manner, where the existing classifiers perform classification in a sequential manner which leads to poor classification of documents. The existing classifiers were designed to fit only for a specific type of dataset either with specific size, where ARC is designed to fit for document dataset with any size. For evaluating the performance of classifiers, this research work has chosen ACM Document collection dataset, Reuters-21578, NBA Input document collection dataset of a B-School which holds 3506, 21578, and 1256 documents respectively. The results shows that ARC is having better performance in terms of Classification Accuracy and F-Measure, than baseline classifiers.