franciscoliu/customer_churning

Customer attrition or “churning” (i.e., more and more customers are leaving their credit card services, such as canceling the cards) is an important problem faced by credit card companies. In this project, you are given a dataset of 10,000 customers with 18 attributes/features, such as, age, salary, marital status, credit card limit, credit card category, and so on. The goal is to build a model that can accurately predict churning customers and help understand the reason behind the scene. The 5-fold cross-validation results will be used in evaluating the performance.

User_Instruction

What preprocess method to use

We have two preprocessing methods. For preprocessing method #1, we replace the unknown data point with the mean value of the known data of that feature. For preprocessing method #2, we replace unknown data using IterativeImputer. The default method is method 1. However, if you want to try the second method, you can change "X = X1" to "X = X2", "y = y1" to "y = y2" and "feature_mapping1" to "feature_mapping2".

What classifier to use

We have 6 classifiers in the project. The default classifier is xgboost. If you would like to see the behavior of the other classifiers, you can uncomment the corresponding classifier in "name" and "classifiers" lists.

Note: It takes approximately 50 minutes to run all the classifiers at the same time.