Create an India credit risk(default) model, using the data provided in the spreadsheet raw-data.xlsx, and validate it on validation_data.xlsx. Please use the logistic regression framework to develop the credit default model.
Data description – Please direct them to the video – Default Risk Prediction. After removing variables for multicollinearity, we should try to take at least one variable for creating the model from each of the 4 factors namely –
4) Company’s size
For Default Risk Estimation, all the variables are bifurcated in different buckets in the categories tab in raw_data file.
Creation of new variables – This is an important step in the project as the company which is the biggest in size, will also have bigger asset size, cash flows, etc. (Hint: We need to think in terms of ratios – Equity to asset ratio, debt to equity ratio, etc)
Dependent variable – We need to create a default variable which should take the value of 1 when net worth is negative & 0 when net worth is positive.
Validation Dataset – We need to build the model on the raw dataset and check the model performance measures on the validation dataset.
Please find attached the files to be referred.
Please note the following:
- You have to submit 2 files :
- Business Report not exceeding 3000 words. In this, you need to submit all the answers to all the questions in a sequential manner. Your answer should include detailed explanations & inferences to all the questions. Your report should not be filled with codes.
- R code file: This is a must and will be used for reference.
1. Outlier Treatment – Outlier Treatment
2. Missing Value Treatment
3. New Variables Creation (One ration for profitability, leverage, liquidity and company’s size each )
4. Check for multicollinearity
5. Univariate & bivariate analysis
6. Build Logistic Regression Model on most important variables
7. Analyze coefficient & their signs
8. Predict accuracy of model on dev and validation datasets
9. Sort the data in descending order based on probability of default and then divide into 10 deciles based on probability & check how well the model has performed