•1 min read•from Data Science
Clients clustering: How would you procede for adding other than rfm variables to kmeans?
Our take
When enhancing your RFM clustering with additional variables, consider the relationship between the new data and your existing clusters. You could integrate the change variables and other factors into the same K-means model, ensuring a comprehensive view of your clients. Alternatively, applying K-means separately within each RFM cluster might reveal nuanced insights specific to those segments. To evaluate the effectiveness of added variables, use validation techniques like silhouette scores or the elbow method to assess cluster cohesion.
I have my RFM clustering. I want to add:
change variables: ratio q1 to year, ratio q2 to q1, ration q3 to q2, S1 to S2...
other variables: returns of products, channel ( web, store..), buying by card or cash, navigation data on the web...
Would you do that in the same kmeans and mix with rfm variables? or on each rfm cluster do another kmeans with these variable? or a totally separate clustering since different data ( web navigation)? how to know if it is good to add the variable or not? is it bad to do many close variables like ratio q2 to q1, ration q3 to q2? how would you procede, validate...?
[link] [comments]
Read on the original site
Open the publisher's page for the full experience