UTFacultiesEEMCSDisciplines & departmentsSCSEducationAssignmentsFinished AssignmentsFinished Master AssignmentsImplementing Ensemble Machine Learning Models for Anomaly Detection in Credit Risk Data Quality

Implementing Ensemble Machine Learning Models for Anomaly Detection in Credit Risk Data Quality

MASTER Assignment

Implementing Ensemble Machine Learning Models for Anomaly Detection in Credit Risk Data Quality

Type : Master M-CS

Period: October, 2024 - March, 2025

Student : Mădăras, A.A. (Alexandru, Student CS)

Date Final project: March 21, 2025

Thesis

Supervisors:

Abstract:

This thesis explores and compares various anomaly detection methods to enhance data quality issues detection in credit risk analysis for financial institutions. Although various taxonomies exist for anomaly detection methods, this thesis consolidates them into a unified classification based on an extensive literature review. Using this classification, the thesis selects and compares relevant anomaly detection methods from the literature through experiments conducted on a data set from a top European financial institution, ING Bank. These experiments employ parallel (bagging) ensemble learning techniques and leverage Explainable AI to explain the models’ decisions to identify specific rows as potential data quality issues. Keywords: Data quality, credit risk analysis, data quality, anomaly detection, artificial intelligence, ensemble learning, Explainable AI.