Conducted a comparative machine learning study to predict sustainable fashion adoption among Gen Z consumers in Bangladesh. Evaluated Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, and SVM using systematic feature engineering and cross-validation. Gradient Boosting achieved the best performance (85% accuracy, 75% F1-score, 90% AUC). Applied SHAP-based explainability to identify economic affordability and promotional dependency as the most influential behavioral factors, highlighting affordability as a key barrier to adoption.

Developed an integrative machine learning pipeline combining Random Forest and LSTM models to predict respiratory irritation from biomass combustion exposure. Fused environmental metrics (PM2.5, NOx, VOCs) with clinical biomarkers for real-time health risk stratification. Identified nitrogen oxides as the most influential predictor, revealing mechanistic links between combustion inefficiency and inflammatory response.