Visualization and Explainable AI (XAI) Techniques in Check Fraud Detection

Check fraud continues to be a substantial challenge in the financial sector, involving the unauthorized use of checks for illegal fund acquisition. To address the issue of fraud detection, specific Artificial Intelligence models, particularly Convolutional Neural Networks (CNNs), have been increasingly applied to classify checks as fraudulent or legitimate. However, there is a crucial aspect of these models usually overlooked, which is the ability to interpret and understand the decisions made by the model. The ability to interpret why a check is labelled as fraudulent or legitimate is highly important, helping to understand the basis for a decision which increases transparency and trustfulness. In this project, two different AI techniques are applied to our model to generate the heatmaps of the final classified output. They reveal which part of the check is most influential in the model’s decision-making process, allowing Verafin to enhance the accuracy and fairness of its fraud detection systems.

Faculty Supervisor:

Sarah Power;Reza Shahidi;Reza Shahidi

Student:

Partner:

NASDAQ Canada Inc

Discipline:

Engineering

Sector:

Finance and Insurance; Artificial Intelligence

University:

Memorial University of Newfoundland

Program: