0
$\begingroup$

I am involving in a project of customer churn prediction. I have 10000 customers data. All the data are unlabeled data. Now my question is can I perform CART analysis with unlabeled data?

$\endgroup$
5
  • 2
    $\begingroup$ I don't think that is possible. Could you tell us more about your problem? If you are trying to predict "churn" then the data are labeled (by churn). $\endgroup$
    – Peter Flom
    Commented Jul 9, 2013 at 10:57
  • $\begingroup$ @Peter: I have the customer transaction data of a anonymous bank. The data contains customer's deposit and withdraw details(cr amount, dr amount, transaction timestamp, account type) along with customer profile. The raw data are not labeled. And I am thinking usign CART for this. $\endgroup$ Commented Jul 9, 2013 at 11:55
  • $\begingroup$ CART demands labels on the variable you are trying to predict/explain. $\endgroup$
    – Peter Flom
    Commented Jul 9, 2013 at 12:04
  • $\begingroup$ @Peter: Could you suggest the good algorithm for churn prediction? $\endgroup$ Commented Jul 9, 2013 at 12:34
  • $\begingroup$ Sorry but I have no idea. CART might be good. I still think your data are labeled. If you don't know which customer is which, then you can't do anything that I can see $\endgroup$
    – Peter Flom
    Commented Jul 9, 2013 at 18:38

1 Answer 1

1
$\begingroup$

As Peter Flom notes, you can't use CART on unlabeled data. Nor any other supervised learning technique. All you can do is to cluster your data. Which will not help in predicting churn, if churn doesn't enter the model in some way.

However, if you have transaction, deposit and withdrawal data, then you can label your data. Anyone with a zero balance and no transactions for a while has almost certainly left your bank, i.e., they have churned. (You would need to operationalize "a while", which may depend on their transaction frequency prior to their last transaction.)

After you have labeled your data in this way, you can use CART, or any other supervised learning technique.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.