Variance-Inter-Quartile Range Hierarchical Clustering Method with Applications (VIQR)

Clustering is the task of grouping a set of objects in such a way that objects in the same group (cluster) are more similar  to each other than to those in other groups (clusters).  In this paper, a new procedure of linkage hierarchical clustering method (Variance-inter-quartile Range Hierarchical clustering method) was developed to reduce the effect of extreme values in the classification of groups into clusters. The new proposed method makes use of the variance to get the variance matrix and inter-quartile range for the construction of the dendogram which is used for the comparative analysis of the newly proposed method and the existing methods. A performance evaluation was carried out using the visual dendogram inspection and silhouette score on the newly proposed method VIQR (Variance-inter-quartile Range Hierarchical clustering method) and the existing methods (Single and Complete linkage). From the result, using visual dendogram inspection, It was observed that VIQR Hierarchical clustering method performs better than the existing methods for the classification of objects as it has similar dendogram with fewer steps in the classification procedure. Using Shilouette score as a method of validation of the clusters, from the result, the score for the proposed method is 0.53 which is higher than the scores of the existing methods; single linkage (0.23) and complete linkage linkage (0.30). In summary, VIQR is more efficient and robust classification method in terms of shorter algorithm, better control of extreme values, and less complexity. Therefore the VIQR should be adopted  as a better method for classification of observations or variables especially when dealing with problems that have to do with classification.