Post by Maryam Jabbar, CIMA Dip MA

As part of the JPMorganChase Job Simulation, I worked on a Credit Risk task to estimate the Probability of Default (PD) for loan borrowers. Rather than stopping there, I wanted to understand what actually drives default risk. So I expanded the project by: • Engineering features such as Debt-to-Income (DTI) and Loan-to-Income (LTI) ratios. • Performing exploratory data analysis to identify borrower trends. • Analysing logistic regression coefficients to interpret the key drivers of default. • Calculating Expected Loss (EL) and creating business-focused visualisations to communicate the findings. Some key insights: 📊 Borrowers who defaulted had an average DTI of 27.2%, compared with 9.0% for non-defaulters. So higher DTI ratio means a higher financial risk 📉 Longer employment history was associated with lower default risk. 📈 Credit Lines Outstanding and Debt-to-Income Ratio emerged as the strongest drivers of default, while higher income, stronger FICO scores, and longer employment history reduced risk. This project reinforced an important lesson: a model predicts outcomes, but analysis explains them. Turning model outputs into insights that stakeholders can understand is just as valuable as building the model itself. I'm still early in my journey in credit risk and would genuinely love to hear how professionals in this field approach projects like this. Are there additional analyses, validation techniques, or business considerations you would typically include beyond building the model? #CreditRisk #RiskAnalytics #Finance #DataAnalytics #Python #MachineLearning #LogisticRegression #CIMA #JPMorganChase