Nondela129/Fraud-Detection-SAS

Fraud detection project using SAS with logistic regression, ensemble models, and visualizations for South African banking context. # Fraud Detection Project (SAS) This project applies SAS analytics to detect credit card fraud, reflecting challenges in the South African banking sector. ## Key Features - Data preprocessing and feature engineering in SAS - Logistic regression baseline with class weighting - Advanced models: Random Forests (HPFOREST) and Gradient Boosting (GRADBOOST) - Evaluation metrics: Accuracy 99.6%, Sensitivity 87.2%, Specificity 99.6% - Visualizations: Fraud distribution, transaction patterns, ROC curve ## Repository Structure - `code/` → SAS scripts for preprocessing, modeling, evaluation - `visuals/` → PNG plots (Fraud counts, Amount histogram, Boxplot, Scatter, ROC curve) - `docs/` → Project report with captions and conclusions - `data/` → Kaggle dataset reference ## Results Fraudulent transactions cluster at unusual hours and higher amounts. The SAS models achieved strong performance, demonstrating practical value for real-time fraud detection. ## References - Kaggle Credit Card Fraud Dataset: kaggle.com - SABRIC Annual Crime Statistics Report - EngineerIT & RCS Group articles on AI and fraud