Classifying text documents using unconventional representation

Classification of text documents is one of the most common themes in the field of machine learning. Although a text document expresses a wide range of information, but it lacks the imposed structure of tradition database. Thus, unstructured data, particularly free running text data has to be transferred into a structured data. Hence, in this paper we represent the text document unconventionally by making use of symbolic data analysis concepts. We propose a new method of representing documents based on clustering of term frequency vectors. Term frequency vectors of each cluster are used to form a symbolic representation by the use of Mean and Standard Deviation. Further, term frequency vectors are used in the form a interval valued features. To cluster the term frequency vectors, we make use of Single Linkage, Complete Linkage, Average Linkage, K-Means and Fuzzy C-Means clustering algorithms. To corroborate the efficacy of the proposed model we conducted extensive experimentations on standard datasets like 20 Newsgroup Large, 20 Mini Newsgroup, Vehicles Wikipedia datasets and our own created datasets like Google Newsgroup and Research Article Abstracts. Experimental results reveal that the proposed model gives better results when compared to the state of the art techniques. In addition, as the method is based on a simple matching scheme, it requires a negligible time.