I am thinking to improve the imbalanced dataset for my churning model, as most people recommend like over/under sampling. I am wondering if using past customer churn data would be helpful. Say that I am now collecting data for the past 12 months only to start with, and for this purpose I also collect customer churn data from past 12-36 months. Any feedback would be appreciated.
Thank you
Topic churn
Category Data Science
Sorry if I have not clearly understood your question. But churn data usually will have imbalance for which you should use some method like SMOTE or its variants like SMOTENC, SMOTEN, KmeansSMOTE etc. which will create synthetic samples of minority class, not just copying. Of course you can collect past data but still, the target variables will be of imbalanced and it being a classification problem, it is not relevant on how much past data you collect. Only the total number of samples and the imbalance in the target classes
Geeks Mental is a community that publishes articles and tutorials about Web, Android, Data Science, new techniques and Linux security.