A cluster analysis is performed inorder to identify groups of articles that share similar characteristics and therefore form. Hierarchal cluster analysis attempts to identify relatively homogenous groups of cases based on selected characteristics, using an algorithm that starts with each case (variable) in a separate cluster and combines clusters until only one is left. This, this technique allows for an iterative process in order to determine the optimal number of clusters in terms of the degree of homogeneity desired.
Within hierarchal cluster analysis, two basic cluster hierarchical clustering procedures can be differentiated; agglomerative and divisive. Agglomerative starts by defining each object or article as a single cluster and combines these to new clusters until eventually all objects are grouped into one large cluster. Divisive, on the other hand, proceeds in the opposite direction and seeks to divide one large cluster into smaller groups.
One can use the hierarchical clustering in the supply chain in terms of clustering the warehouses based on certain characteristics such as size, type of products stored etc. The agglomerative clustering technique can be used to cluster warehouses. At the start of the process, each warehouse can be considered as a separate cluster. These clusters are then grouped together according to similarity until one cluster remains. The point needs to be determined where the two warehouses/articles are sufficiently similar to be grouped together. This is done by amalgamation or linkage rules which are algorithms targeted at the combination of objects in a data set.
There are different algorithms that can be applied as an amalgamation rule for hierarchical cluster analysis. The most important algorithms are
· Single linkage or nearest neighbor: this algorithm determines the distance between two clusters by using the distance of the two nearest articles (nearest neighboring warehouses) in the different clusters. In essence, the result constitutes clusters that tend to represent long chains and the clusters at the two ends of the chain are those that are least likely to the others
· Complete linkage or furthest neighbor: this algorithm determines the distance between two clusters by means of the greatest distance (furthest neighboring warehouses) between any two articles in the different clusters
In order to determine the optimum result in supply chain management that is characterized by high level of stability, both these algorithms were computed. In the first step, the nearest neighboring algorithm was used to determine outliers at the two ends of the chain of warehousing clusters which will then be excluded from further analysis. Then the farthest neighbor analysis is calculated to exclude the warehouses which are very close to each other. The result would be warehouses which are similar in characteristics and are segregated based on distance.
Operatons – Group3