Fraud Detection Pipeline

n end-to-end data preprocessing pipeline built in Python for credit card fraud detection on a real-world Kaggle dataset of 284,000+ transactions. Covers full EDA, duplicate removal, IQR-based outlier clipping, log-transform and time-based feature engineering, stratified train/test splitting, a reusable scikit-learn preprocessing pipeline, and SMOTE oversampling to address severe class imbalance. Validated with a Logistic Regression baseline model.