Text document categorization using modified K-Means clustering algorithm

Abstract

The volume of the information that is to be managed is increasing at exponential pace. The challenge arises how to manage this large data effectively. There are many parameters on which the performance of such a system can be measured such as time to retrieve the data, similarity of documents placed in same cluster etc. The paper presents an approach for auto-document categorization using a modified k-means. The proposed methodology has been tested on three different data sets. Experimental findings suggest that proposed methodology is accurate and robust for creating accurate clusters of documents. The proposed methodology uses cosine similarity measure and a fuzzy k-means clustering approach to yield the results very fast and accurately.