Discovery of Knowledge in databases requires not only the growth of novel mining techniques but also fair and comparable quality assessment based on objective evaluation measures. Quality measurement of a clustering algorithm has shown to be as important as the algorithm itself and new challenges for the clustering algorithms as well as for their evaluation measures. In many cases, external evaluation measures were exclusively used for validating clustering algorithms. Most of the applications unable to produce external validation but it requires a ground truth. Subspace clustering techniques can create clusters which are useful in different subsets of the full feature space. Especially in the enhanced subspace clustering of internal validation is effective and realistic in nature. Silhouette coefficient index (SCI) is one of the most famous and efficient internal measures in the evaluation of clustering validity. This article analyzes the importance and execution of Silhouette Coefficient internal measure. In particular we apply this measure to carefully synthesized enhanced applications on evolving density based subspace clustering algorithms. The SCI performs the best in coping with the cohesion among the members of the cluster and separation of the cluster to its closest cluster in subspace clusters of high dimensional data. It also performs the best for density based ones. The real world and synthetic data sets proved the effectiveness of Silhouette coefficient. A number of experimental results show that different from the case with external measures on high dimensional data and validate better clusters of enhanced subspace clustering model called ENSUBCLU than other EDENCOS, INSCY, SUBCLU subspace clustering algorithms.
Volume 11 | Issue 1
Pages: 321-328