When: Wednesday, October 22, 2025
Where: RB 104
Time: 3:00 PM
Title: Advanced Fraud Detection Methods in Imbalanced Data Environments
Presented By: Azadeh Abdollah Zadeh, Assistant Teaching Professor of Computer Science
Abstract: Fraud detection in large-scale financial and healthcare systems is challenging due to severe class imbalance and limited labeled data. This study explores One-Class Classification (OCC) methods for identifying fraudulent activities in highly imbalanced datasets. Using Credit Card and Medicare claims data, the One-Class SVM (OCSVM), One-Class Gaussian Mixture Model (OCGMM), and One-Class Adversarial Network (OCAN) are evaluated against traditional binary classifiers such as CatBoost and Random Forest. Results show that while binary classifiers outperform OCC models when both class labels are available, OCGMM with Sigmoid calibration achieves robust performance in one-class settings, effectively managing imbalance and data complexity. The findings demonstrate that OCC remains a practical solution when minority labels are scarce, offering valuable insights for advanced fraud detection in big data environments.
Bio: Azadeh Abdollah Zadeh earned her PhD in Computer Engineering from Florida Atlantic University. Her research focuses on machine learning, data mining, and bioinformatics, with publications on topics such as one-class classification, big data analytics, deep learning for facial emotion recognition, gene analysis in Parkinson’s disease, and fraud detection. She is particularly interested in improving classification techniques to enhance accuracy and efficiency in analyzing complex datasets and comparing the performance of one-class and binary classification systems in real-world applications.
Light refreshments will be served