Process using k means on customer segmentation

tonyboy9

See my screen shot 'Process using k means.'

I ran it and to my surprise, it executed.

Result was the screen shot 'K means cluster model.'
What is this telling me?

I clicked on Plot in the side bar, got the screen shot 
'K means cluster model plot.' What is this telling me?

This is the problem statement accompanying the data set.

Any help is appreciated.

jacobcybulski

Normally, to get good k-Means clustering you need to normalise the attributes, e.g. to the interval 0..1. If one attribute (such as D) is in the range of 0..475 and a dummy encoded attribute is in the range 0..1, the cluster chart is very difficult to interpret. Also you cluster model will be biased towards those large-valued attributes, as their values will dominate your distance measurements, which are at the core of k-Means (especially when you rely on Euclidean measures for numeric attributes, which I think is default). Also note that the presence of a large number of dummy variables will make difficult to interpret the cluster chart.
Jacob 

jacobcybulski

Normally, to get good k-Means clustering you need to normalise the attributes, e.g. to the interval 0..1. If one attribute (such as D) is in the range of 0..475 and a dummy encoded attribute is in the range 0..1, the cluster chart is very difficult to interpret. Also you cluster model will be biased towards those large-valued attributes, as their values will dominate your distance measurements, which are at the core of k-Means (especially when you rely on Euclidean measures for numeric attributes, which I think is default). Also note that the presence of a large number of dummy variables will make difficult to interpret the cluster chart.
Jacob