Improvement in K-Means Clustering for Information Retrieval