Regression analysis is a powerful tool in social science research.

It lets researchers make intuitive statements about social phenomena — for example, “a one-unit increase in poverty rate is associated with an average increase of β in crime rate.”

But as society grows more complex, more and more phenomena resist explanation through linear relationships alone.

Consider life satisfaction: it does not increase proportionally with income. It rises up to a point, then levels off. Variables like education, income, leisure activity, and health tend to move together, making it difficult to isolate each one’s independent contribution to life satisfaction. And when a particular individual or small group shows unusually high or low life satisfaction, the regression coefficients derived from the full sample cannot explain that specific case.

Machine learning (ML) and explainable AI (XAI) address these limitations. ML can flexibly capture nonlinear patterns and variable interactions, and XAI makes it possible to quantify how much each variable contributed — both across the full dataset and for individual cases.

This tutorial walks through the full workflow step by step: building a predictive model with machine learning, identifying which independent variables contributed most to the predictions, and examining individual cases in detail that regression alone could not explain.

☞ Try for Free

What will you learn in this tutorial?

• How to build a predictive model with machine learning
• How to identify which variables contributed most to predicting the dependent variable
• How to decompose each variable’s contribution to predictions at the individual case level

Who is this tutorial for?

• Researchers who have used regression but need to account for nonlinear relationships or variable interactions
• Anyone who wants to explain why a model made a specific prediction
• Anyone who wants to go beyond average patterns and understand specific cases in depth

NetMiner 5 features used in this tutorial

• Machine Learning > Learn (Classical) — supervised learning
• Machine Learning > XAI > SHAP