Comparative Study of Single Linkage, Complete Linkage, and Ward Method of Agglomerative Clustering

Clustering is the process of grouping the datasets into various clusters in such a way which leads to maximum inter-cluster dissimilarity but maximum intra-cluster similarity. Clustering has a wise application field like data concept construction, simplification, pattern recognition etc. Clustering method is broadly divided in two groups, one is hierarchical and other one is partitioning. In hierarchical clustering method the hierarchy of clusters is shown by splitting and combining them at different levels whereas in partitioning method partitions are formed and evaluated based on some criteria. Thus clustering algorithms chosen need to be efficient. The detail about clustering and hierarchical clustering methods is explained in our previous paper[a]. This paper focuses on the real life application of clustering As well as a comparative study over the different linkage techniques or methods used to calculate the decision factor for merging of clusters at any level. In this paper we have implemented hierarchical agglomerative clustering over a real time shopping data and tried to gain some useful insights from them. We have also compared the result of different metrics like “ward”, “single linkage” “complete linkage” etc. and analyzed how the result varies