Solo Project Presentation

Problem and Objectives

The Modern Fraud Landscape: Fraud has evolved beyond stolen credit cards into behavioral anomalies, linguistic fraud (fake reviews), and automated "flash transactions."

Limitations of Traditional Systems: Traditional rule-based systems relying on human-written "IF-THEN" logic and geographic triggers struggle with speed and complexity.

Research Goals: The project aims for architectural diversity, high precision, scalability for real-time detection, and optimization for imbalanced datasets.

Methodology and Models

The presentation details three primary algorithmic approaches:

XGBoost: Focuses on sequential error correction (boosting) and advanced regularization to handle noise and overfitting.

LightGBM: Utilizes leaf-wise tree growth and Gradient-based One-Side Sampling (GOSS) for faster training and lower memory usage.

Artificial Neural Networks (ANN): Explores architectural layers to model behavioral biometrics and contextual patterns, incorporating SMOTE to handle data imbalance.