We calculate PD,LGD,EAD and Expected loss using logistic and beta regressions.
Notebook 1: Training dataset preparation (Feature engineering - Continuous and Discrete dataset)
Notebook 2: Test dataset preparation
Notebook 3: Calculate Probability of Default (POD) and Credit score card preparation
Notebook 4: Monitor the POD model
Notebook 5: Calculate LGD (Loss given dataset) and EAD (Exposure at Default)for individual customers
Notebook 6: Compute Expected Loss for individual customers
Dataset (Source:Lending Club)
- loan_data_2007_2014 (Train dataset)
- loan_data_2015 (Validation dataset)
- loan_data_2007_2014_preprocessed (Preprocessed dataset)
Techniques covered in the model building:
-
Weight of evidence
-
Information value
-
Fine classing
-
Coarse classing
-
Linear regression
-
Logistic regression
-
Area Under the Curve
-
Receiver Operating Characteristic Curve
-
Gini Coefficient
-
Kolmogorov-Smirnov
-
Assessing Population Stability
-
Maintaining a model