0
$\begingroup$

I'm trying to classify churn (1 or 0) for a user based on day-to-day time series and activity levels, something like this:

User     Date      Activity Churn
1    09-01-2018       35      0
1    09-02-2018       17      0
2    09-01-2018        0      1
2    09-02-2018       13      1
etc...

Could someone give a few pointers on how this would be setup in Keras to predict the churn value?

$\endgroup$
0
$\begingroup$

you want to treat the date as a cyclical feature to capture patterns in the time stamp.

http://blog.davidkaleko.com/feature-engineering-cyclical-features.html

Keras + SK learn work nicely together. here's an example in my GitHub to extract patterns from the time stamp (just adjust to your needs and replace my Decision Tree with Keras code).

https://github.com/FrancoSwiss/Fraud_Detection/blob/master/Kaggle%20Fraud%20(1).ipynbenter image description here]1

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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