I have a problem with my current project. The goal is to create clusters of customers with homogenous properties. So, each cluster should contain a group of customers with similar attributes.
I already used k-means clustering with k = 5 and 3 variables, to identify different segments of my customers, but the result seems not appropriate for me.
In my opinion, one variable - ordering time during the day (midday or evening) - is the most important variable. When I apply the k-means clustering only on the ordering time during the day I can identify two segments/clusters. Now I want to conduct a further k-means clustering on the two already identified clusters. That means, I want to split the clustering process into two parts to get approximately 5 clusters. Can you explain to me whether this approach is a good idea?
