The sustainable fashion adoption behavior of Gen Z requires predictive frameworks that are strong and cannot be predicted using the conventional survey-based designs. In this paper, a comparative machine learning analysis is carried out to forecast sustainability-oriented buying behavior using five algorithms: Logistic Regression, Decision Tree, Random Forest, Gradient Boosting and Support Vector Machines, among the Gen Z consumers in Bangladesh. The study determines the economic accessibility and promotional dependency as the most predictive behavioral factors through systematic feature engineering, crossvalidation procedures, and describe the outcomes through the SHAP-based explainability analysis. The highest performance of Gradient Boosting was 85 percent accuracy, 75 percent F1-score, and 90 percent AUC, which showed better ability to describe nonlinear behavioral interactions. The interpretability of the framework indicates that affordability barriers are the key barriers to adoption, which daunt attitude-based assumptions that are common with sustainability research. Such a comparative ML strategy creates methodological bases of sustainable consumption transition policy interventions in emerging economies that bridge theoretical bases of behavior modeling with algorithm prediction facilities.