Explainable Machine Learning for Modeling Online Impulse-Buying Tendency: A Generational Comparison of Gen Z and Millennials

Abstract

Online impulse-buying tendency is widely studied through explanatory models, but less is known about how well it can be predicted across generational cohorts. This study analyzed 540 adult online shoppers in India (299 Gen Z; 241 Millennials) using a nine-item buying-impulsiveness measure. Internal consistency was good (Cronbach’s α = .820), and sampling adequacy was high (KMO = .903), and cohort means did not differ significantly (p = .202). Five regression algorithms were evaluated with nested cross-validation. Ridge regression performed best in the pooled sample (R² = .155; RMSE = .647; MAE = .524), while a behavioral-only feature set improved R² to .171. SHAP attribution identified discount responsiveness, unplanned browsing, and recommendation exploration as the leading predictive signals. Cohort-specific explanations showed similar core predictors but different secondary rankings. The results favor parsimonious, interpretable prediction over unnecessary model complexity in survey-based consumer analytics.