EXPLORE PUBLICATIONS BY COUNTRIES

Text Clustering, Partitional Approach, Initial Centroids, Similarity Measures, Cluster Accuracy.

Text document clustering is gaining popularity in the knowledge discovery field for effectively navigating, browsing and organizing large amounts of textual information into a small number of meaningful clusters. Text mining is a semi-automated process of extracting knowledge from voluminous unstructured data. A widely studied data mining problem in the text domain is clustering. Clustering is an unsupervised learning method that aims to find groups of similar objects in the data with respect to some predefined criterion. In this work we propose a variant method for finding initial centroids. The initial centroids are chosen by using farthest neighbors. For the partitioning based clustering algorithms traditionally the initial centroids are chosen randomly but in the proposed method the initial centroids are chosen by using farthest neighbors. The accuracy of the clusters and efficiency of the partition based clustering algorithms depend on the initial centroids chosen. In the experiment, kmeans algorithm is applied and the initial centroids for kmeans are chosen by using farthest neighbors. Our experimental results shows the accuracy of the clusters and efficiency of the kmeans algorithm is improved compared to the traditional way of choosing initial centroids.