Optimizing Fraud Detection with XGBoost and CatBoost for Social Media Profiles and Payment Systems

This study assesses advanced machine learning algorithms, namely XGboost and Gradient Boosting Machine (GBM), for fraud detection in social media profiles and payment systems. It focuses on two very different types of datasets: one dealing with fake social media accounts and another with fraudulent payment records. Consequently, the study aims toward establishing a solid methodology for the detection of outliers. The main tasks comprise deep cleaning and preprocessing, feature engineering, and fine-tuning of the algorithm to present optimal detection accuracy and performance. Results exhibit the usability of ensemble learning techniques in handling various fraud detection conditions for providing better security and reliability across the respective platforms.