SQL Server Data Mining—Understanding the Composition of Clusters

Microsoft .NET Framework, ASP.NET, Visual C# (CSharp, C Sharp, C-Sharp) Developer Training, Visual Studio

Understanding the Composition of Clusters

The Cluster Profiles view shows all the clusters that were identified as columns, and each attribute that you selected for your model as the rows, as shown in Figure 10-7. Looking first at Cluster 1, we can see that all the customers in the group have an Internet Purchaser attribute of False, as well as an Internet Visitor of False. So, the mining algorithm has grouped customers together who have never visited the site or purchased anything online—all their purchases have been at a physical store. Note that we can come to this rather useful conclusion only because we understand the underlying business, which is a key point about data mining.

Figure 10-7 Cluster profiles

To give Cluster 1 the more sensible name of Store-Only Buyers, right-click the cluster name and select Rename Cluster. So, we now have a single cluster identified; what about the others? If you look at the next column, you can see that Cluster 2 differs from Store-Only Buyers in that all the customers in the cluster have actually visited the site, but they just haven't made any purchases online yet. We can call this cluster Browsers because they are customers who are (so far) using the site for information gathering only.

Cluster 6 contains visitors who have also made a purchase, but if we look closely at the Months Internet Purchaser and Months Internet User attributes, we learn that they are all relative newcomers to our site—all of them have been visitors and purchasers for between zero and three months (they are "Newbies"). We can continue the process of looking at each cluster, but the rest of the clusters are not quite so clear-cut, so we need a better tool for differentiating between them.