Our cluster analysis add-in for Caliper Maptitude, MPCluster for Maptitude has a new version ready for testing. Two significant new features have been added. First, it is now possible to define ‘fixed’ cluster centers for the K-Means algorithm. For example, you might be using MPCluster to determine the ideal locations for new warehouses, but already have some operational warehouses. You could import these warehouse locations as fixed cluster centers. MPCluster will always include these fixed locations in the output, although they might not necessarily meet your constraints. For example, there might not be enough customers in their vicinity.
Second, boundaries drawn for clusters derived with the Hierarchical algorithm are now ‘concave hulls’ (aka Alpha Shapes). Unlike the K-Means, the Hierarchical algorithm can produce concave clusters. However, MPCluster 1.1 will draw the boundaries as convex shapes which are quick and simple but can appear to overlap neighboring clusters. This has been fixed with the new ‘concave hull’ boundaries. The images below show a couple of examples.
Let me know if you would like to test the latest beta of v1.2. If you are new to MPCluster, we recommend you familiarize yourself with the released version of MPCluster and Maptitude first.
The current plan is to release v1.2 in January, and it will be a free upgrade for existing licensed users of MPCluster for Maptitude.
Update, 18th December: Another new feature has been added: “Overlays”. These are similar to Maptitude’s Overlays and calculate aggregate statistics (sum and/or count) of a point layer for each cluster boundary shape. For example, you could count the number of customers in each cluster, or sum up the total sales for each cluster. The results are written to columns in the cluster boundary shape. Let me know if you would like to test this.
It is likely that this new ‘overlay’ feature and the fixed clusters (announced above) will require a new ‘Professional’ license. Release date is likely to be late January or February.