End-to-end credit card fraud detection pipeline with preprocessing, feature scaling, and class imbalance handling via undersampling and SMOTE. Trains Logistic Regression and Decision Tree models, evaluates with key metrics, and exports the final model using joblib for manual transaction classification.