End-to-End Fraud Detection System

Developed a full-stack machine learning system for credit card fraud detection.

● Handled class imbalance using oversampling, undersampling, and hybrid resampling strategies

● Optimized models via GridSearchCV for hyperparameter tuning.

● Implemented a stacking ensemble (RandomForest, XGBoost, CatBoost + Logistic Regression)

achieving 88.7% Recall and 85.6% F1-score on unseen test data.

● Deployed as a Flask web app with support for single transaction prediction and batch CSV uploads.

● Tech Stack: Flask, Scikit-learn, XGBoost, CatBoost, Pandas, NumPy, Matplotlib, Seaborn.