Different similarity measures for text classification using KNN

Abstract

Present days humans are associated with many electronic gadgets which generate large amount of data on regular basis. The sole purpose of generated data was to meet the immediate needs and no attempt in organizing the data for later efficient retrieval was attempted. Over the period of time, the data generated became voluminous, this paper attempts to classify the huge data into different categories for easy retrieval. We have many techniques to classify the data which exists in the structured format, but not much work has been addressed when the data is available in textual form. In the present paper an attempt to classify the textual data based on its content is explored. The paper explores the process of building multi-classifier model for textual data. In the process of designing the model the K-Nearest Neighbour paradigm was employed, which has given encouraging results. The paper also attempts to explore different similarity measures, different feature selection techniques in the process of designing textual multi-classification. © 2011 IEEE.