Subscribe to DSC Newsletter

Hey, I wanted to ask that what association test should I perform while doing churn analysis(for a internet service provider) when the available data is categorical, like, sales channel, warranty name, model, etc. also which machine learning method would be best- decision trees, logistic regression, clustering or nearest-neighbour search. Actually I am a student and is required to submit a case study on the churn analysis with the available data of about 20000 customers to get a call for final interview and I have no experience this is my first. 

Views: 961

Reply to This

Replies to This Discussion

Hi Dhruv --

Fortunately there are a number of tools on market targeted at individual data practitioners that can help you perform churn analyses.  We at BigML have published a few models on our gallery -- this one here is a good example. 


Hi Dhruv, first you need to define what is churn. It can be an actual account closure by the customer, or the customer still active but doesn't use the service anymore. This is very crucial to understand the correct definition. The answer will be depend on the business objective.

Next thing is to determine, the performance window. When do you want to predict the churn to happen? The next month? The next 3 month? Do you want to exclude those occasional churner who happen to close the account but open new account again after new promotion?

Since you have only 20K of data and all of the variables are categorical, you might want to start with CHAID or decision tree, just to get a sense of what's happening. May be CHAID is enough to conclude your analysis, especially if the churn rate turns out to be very small percentage.

Good luck.


On Data Science Central

© 2020   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service