Machine Learning in Ms. Excel

Explain credit risk in banking and how a credit officer uses scoring models to estimate the probability of default for mortgages, credit cards, and other loans.

Learn how to combine data sets from different sources using Excel's look up (vlookup) function to achieve exact matches by client I.D., enabling accurate arrears indicators in banking data.

This lecture explains how to create dummy variables in Excel to quantify credit risk and estimate default probability for regression analyses, including gender and urban area encoding using if formulas.

Learn to code the dependent variable for loan default by selecting a days-in-arrears threshold and a binary indicator, where zero denotes default and one denotes non-default.

Compute the weight of evidence for gender in credit risk using Excel pivot tables, comparing male and female good and bad borrowers to enhance modeling.

Compute weight of evidence for urban versus rural residence using good and bad borrower percentages, compare their predictive power, and apply it to credit scoring and modeling.

Compute the weight of evidence for work sector and residence, sort by impact, and visualize as a line graph to select reference variables for regression analyses and machine learning models.

Apply weight of evidence to relationship status to compare loan repayment patterns, noting married individuals show higher weight, guiding regression analyses with married as the reference category.

Construct the weight of evidence for loan grades, normalize to column totals, and use a, b, and c as regression inputs with d or lower as reference.

Learn to code loan maturity into a binary variable in Excel by converting mixed text and numbers, using text-to-columns, and applying an if formula for 1–12 months.

calculate weight of evidence for continuous maturity data by grouping months into year-based categories, compare good and bad borrower proportions, and select reference variables for regression.

correct the regression setup by selecting the three-year maturity as reference, removing the six-year maturity from training dataset, and using veto cup functions to align continuous-variable data with the model.

Explore how real statistics outputs from logistic regression inform coefficient interpretation, significance, and scoring for credit risk, including variable selection, score card creation, and model validation.