This chapter extends the supervised approach by integrating semi-supervised learning (SSL) methods to enhance predictions of LNP transfection efficiency. An iterative pseudo-labeling strategy is detailed, where the model’s most confident predictions on unlabeled data points are incorporated into successive training cycles, effectively enlarging the training dataset and improving model robustness. The chapter explicitly addresses the challenge of dataset imbalance using the Synthetic Minority Over-sampling Technique (SMOTE), further enhancing predictive accuracy. An advanced gradient boosting model (XGBoost), coupled with rigorous hyperparameter tuning via GridSearchCV, is implemented to achieve superior performance. Results presented demonstrate significant improvements in model metrics (accuracy, precision, recall, ROC-AUC), clearly indicating the utility of SSL techniques and data augmentation methods for robust nanocarrier screening predictions.