Hybrid Hierarchical Clustering (KH): Cluster Assessment via Rand index.

This paper introduces a hybrid hierarchical clustering method, this method has several computational advantages over agglomerative hierarchical clustering approach for it uses centroids rather than raw data points. It reduces the sample space for building the hierarchy and hence requires fewer resources. In order to evaluate the hybrid algorithm, it is compared with the standard algorithm in terms of time and accuracy on generating data with different distributions (i.e., uniform and normal) and Chess data sets from the UCI repository for single hierarchical and average hierarchical with Euclidean and Manhattan distances.in this paper, we Determine the number of clusters by using ratio from 0.1 to 0.9 from the total number of original data. And also, we used the external (Rand index) criteria with the purposes to evaluate the results obtained from hybrid hierarchical clustering and standard hierarchical clustering.

Keywords:

Hierarchical clustering , K-means clustering , Hybrid clustering , External validation