Cluster analysis provides a statistical method whereby grouping is done for items of similar attributes from a set of data and its application is well-known for a wide range of disciplines. So, even in the context of grouping similar locations, Cluster Analysis works very well. Therefore, it is being used successfully as a tool for Supply Chain Optimization. Designing a distribution network requires planning of different transportation routes across different areas or deciding the locations for warehouses. We could look at a map covered with dots representing various locations and draw circles around groups which are close to each other. But this approach can be followed by layman as well. By using Cluster Analysis methods, we can group locations in a systematic way and speed up the process of exploring several different versions of the clusters.
We would use Agglomerative Clustering method for the same. We can plot the Cities/locations with the X and Y axes representing Latitude and Longitude respectively. Then we start forming clusters of cities based on the centroid method as nearest or furthest method may not give suitable results for Supply Chain applications. When we use farthest neighbor method, there may be creation of clusters of unequal sizes. If use the average distance method, there is disparity in no of members in a cluster. For such a case we can use Ward's method, which makes assignments which minimize the deviations within the clusters from the centroid. We get relatively equal-sized clusters in such a case. We can use dendograms to decide on the no of clusters and we can reduce the no of clusters based on it. It can be a very useful tool for network or facility location analysis. Rather than some arbitrary decisions on grouping, Cluster Analysis can contribute the analysis that leads to bottom line savings.
Opetations Group 1