Improved K-MEAN Clustering Approach for Web Usage Mining

In the k means clustering algorithm right value of clusters (k) are initially unknown and effective selections of initial seed are also difficult. In this paper efficient k-means algorithm is proposed and implemented which overcome initial seed problem and unknown number of cluster problem. The algorithm is applied on real BIST server log data and Gaussian dataset to test its accuracy and efficiency. At application level this algorithm may used for efficient knowledge discovery from Web repositories.