Enhancing Credit Card Fraud Detection Accuracy by Optimisation of Anomaly Detection Algorithms and Resampling Techniques

Asmath, Aafreen Shan (2024) Enhancing Credit Card Fraud Detection Accuracy by Optimisation of Anomaly Detection Algorithms and Resampling Techniques. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (691kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Credit card fraud has become a major threat to the financial systems, resulting in losses of billions of dollars and reducing the financial trust. Detection methodologies, such as rule-based systems and manual audits, are no longer effective in catching up with the rapidly changing patterns of fraud, besides the challenges posed by an imbalanced dataset. This work attempts to improve the fraud detection accuracy by combining the anomaly detection algorithms, that is, Isolation Forest and Local Outlier Factor with the state-of-the-art resampling techniques, that is, SMOTE, SMOTE-ENN, and Random Under-sampling. The publicly available credit card transaction dataset is employed within this study through preprocessing and feature engineering to handle class imbalance and the data quality. The methodology combines unsupervised anomaly detection techniques with supervised logistic regression, to compare the performance across metrics such as accuracy, precision, recall, the F1 score, and AUC-ROC. It is revealed by the results that logistic regression using SMOTE-ENN obtains the highest AUC value (0.9364) and recall value (0,94), which is effective at detecting fraud transactions and minimizes false positives. Isolation forest with random under-sampling reflects promise but has precision-specific limitations. This study thus emphasizes on how resampling techniques address class imbalance issues along with highlighting logistic regression using SMOTE-ENN as the best solution for fraud detection. Future studies could dwell upon hybrid models and advanced ensemble techniques to improve fraud detection systems in terms of their accuracy, scalability, and robustness.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Uncontrolled Keywords:	Credit Card Fraud Detection; Anomaly Detection; SMOTE-ENN; Isolation Forest; Machine Learning; Resampling Techniques
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HG Finance > Credit. Debt. Loans. Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	01 Sep 2025 14:16
Last Modified:	01 Sep 2025 14:16
URI:	https://norma.ncirl.ie/id/eprint/8674

Actions (login required)

View Item