A model has been created capable of estimating the creditworthiness of a customer, in order to help the dedicated team understand whether or not to accept the request for the issuance of the credit card. The anonymized data is organized in 2 csv files present in the credit_card_approval folder.
The project follows these steps:
- Initial EDA, in order to analyse potential interesting patterns and anomalies;
- Preprocessing to clean the data for the training phase, defining the target label (not included in the original dataframe) and resampling with different techniques - SMOTE, RUS, ADASYN, SMOTEENN;
- Test different classifiers: Logistic Regression, Decision Tree, Random Forest, SGD, KNN, Gradient Boosting and XGBoost;
- Fine tune the classifiers to find their best hyperparameters through Randomized Search.