Regression and Predictive Modeling

Linear and logistic regression, GLMs, model selection, and predictive analytics.

Key Concepts

• •Simple and multiple linear regression: OLS estimation, interpretation of coefficients
• •Model diagnostics: residual analysis, heteroscedasticity, multicollinearity (VIF)
• •Variable selection: forward, backward, stepwise, and best subsets
• •Logistic regression: odds ratios, classification, and ROC curves
• •Generalized linear models: exponential family, link functions, and deviance
• •Poisson and gamma regression for insurance frequency and severity
• •Regularization: ridge, lasso, and elastic net for high-dimensional data
• •Decision trees: splitting criteria, pruning, and interpretation
• •Random forests and gradient boosting for ensemble predictions
• •Cross-validation: k-fold CV for model selection and tuning

Study Tips

• 1.Start with linear regression fundamentals before moving to GLMs.
• 2.Practice interpreting regression coefficients in actuarial contexts.
• 3.Understand the bias-variance tradeoff and how regularization addresses overfitting.
• 4.Compare GLM and tree-based approaches on the same problem to build intuition.
• 5.Always validate models on holdout data, never just training data.

Related Exam Resources