An Enhanced Hybrid Classification Model using Distance Metric to Detect Financial Fraud

Muddanna Hanumantharaya, Yashwanth (2025) An Enhanced Hybrid Classification Model using Distance Metric to Detect Financial Fraud. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (498kB) | Preview
Preview	PDF (Configuration Manual) Download (273kB) | Preview

Abstract

Credit card fraud refers to the use of information to access funds, often detected by measuring the distance between the known and a new transaction. In the financial domain, accurately identifying fraudulent tractions are critical yet challenging due to suboptimal distance metrics that lead to poor detection of fraudulent patterns, especially when the data is scaled unevenly. This research proposes an Enhanced Hybrid Classification model to detect financial fraud from the distance between transactions. The research contributes sustainable development goals by supporting secure, innovative, and resilient financial structure and promoting strong institutions through enhanced fraud through enhanced fraud preventions (SGD 9 and SGD 16). The proposed model combines a hybrid classification model a Distance based Model and deep learning model. The model is implemented using K-Nearest Neighbors and a Multilayer Perceptron, using Soft-Voting to merge their results. The distance-based model is experimented using Euclidean, Manhattan and Minkowski metrics within the KNN classifier. In this research, the hybrid model which combines the strengths of best distance were selected based on F1 scores, The best distance parameters were tuned using manual and RandomizedSearchCV respectively. The credit-card data with a total of 284,870 records, out of which there are only 492 cases of fraud. The result indicates that, the enhanced model with the best params have achieved highest Accuracy and Recall values. The proposed work will benefit financial institutions, customers by reducing financial losses.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Stynes, Paul UNSPECIFIED
Uncontrolled Keywords:	Euclidean; Manhattan; Minkowski; K-Nearest Neighbors (KNN); Multilayer Perceptron (MLP); Soft-Voting
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence H Social Sciences > HG Finance > Credit. Debt. Loans. H Social Sciences > HV Social pathology. Social and public welfare > Criminology > Crimes and Offences > Cyber Crime Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	02 Jun 2026 11:28
Last Modified:	02 Jun 2026 11:28
URI:	https://norma.ncirl.ie/id/eprint/9333

Actions (login required)

View Item