Predict the probability of Customer Churn using Historical Customer data and Customer Features
Churn Prediction is an important problem that is beneficial for a number of companies. In companies such as Netflix where your major revenue is dependent on customer retention it is very important to identify customers who are most likely to leave/unsubsribe from your service and more over find the factors leading to churn.
This the question we are trying to answer. While there maybe numerous reasons why a customer decides to stop using services. It is possible to find patterns in factors/features of a customer profile to be able to predict churn.
It would be most beneficial to understand that if a group of customers who are depending on a certain service is causing and issue. Predicting the group of users and their common attribute could help triangulate the source of problem and eliminate it before churn.
More at Churn Prediction Notebook
| Model | AUC |
|---|---|
| Neural Network | 0.927761 |
| AdaBoost | 0.920597 |
| Logistic Regression | 0.918806 |
| Naive Bayes | 0.890448 |
| Decision Tree | 0.878209 |
| Random Forest | 0.737612 |
| Gaussian Process | 0.573731 |
| Nearest Neighbors | 0.500000 |
| Linear SVM | 0.500000 |
| RBF SVM | 0.500000 |
| Quadratic Discriminant Analysis | 0.390448 |