3/13/24, 10:04 PM SAP HANA PAL – K-Means Algorithm or How to do Cust...

- SAP Community

Community

SAP Community  Products and Technology  Technology  Technology Blogs by Members  SAP HANA PAL – K-Means Algorithm or How to do Cust...

Technology Blogs by Members

Explore a vibrant mix of technical expertise, industry insights, and tech buzz in member blogs covering SAP products, technology, and events. Get in the
mix!

All community  What are you looking for today?

 Due to SAP Community maintenance, portions of the site will be unavailable. We expect this downtime to last a few
hours.

SAP HANA PAL – K-Means Algorithm or How to do Customer Segmentation for the
Telecommunications Industry

Former Member


‎03-28-2013 3:11 PM

 5 Kudos

https://community.sap.com/t5/technology-blogs-by-members/sap-hana-pal-k-means-algorithm-or-how -to-do-customer-segmentation-for-the/ba-p/12976696/page/2 1/39

3/13/24, 10:04 PM SAP HANA PAL – K-Means Algorithm or How to do Cust... - SAP Community

After all the objects are plotted, the algorithm calculates the distance between them and the ones that are close to each
other are grouped in the same cluster. So if we go back to our previous example we can create 4 different clusters:

Each cluster is associated with a centroid and each point is assigned to the cluster with the closest centroid. The centroid is
the mean of the points in the cluster. The closeness can be measured using:

Manhattan Distance
Euclidean Distance (most commonly used)
Minkowski Distance

Every time a point is assigned to a cluster the centroid is recalculated. This is repeated in multiple iterations until centroids
don’t change anymore (meaning all points have been assigned to a corresponding cluster) or until relatively few points
change clusters. Usually most of the centroid movement happens in the first iterations.

One of the main drawbacks of the K-Means Algorithm is that you need to specify the number of Ks (or clusters) upfront as
an input parameter. Knowing this value is usually very hard, that is why it is important to run quality measurement
functions to check the quality of your clustering. Later in this post we will talk about this.

I came across a very interesting paper that talks about segmentation in the telecommunication industry, so I thought it
would be a very nice use case to demo the K-Means algorithm in HANA (if you are interested in this topic, I very much
recommend reading this paper). These are the steps I followed:
https://community.sap.com/t5/technology-blogs-by-members/sap-hana-pal-k-means-algorithm-or-how -to-do-customer-segmentation-for-the/ba-p/12976696/page/2 3/39