Fraud Detection Analytics: An AI solution for duplicate payment and fraud detection

The Problem

Prior to this initiative, the City relied on manual, disparate checks to identify fraudulent and duplicate payments to external vendors, leaving financial operations vulnerable to evolving fraud tactics. Without an integrated, automated system, the City could not cross-reference transactions—such as a single vendor being paid the same amount via both credit card and purchase order—nor could it systematically apply complex predictive risk scoring to proactively flag and rank high-risk anomalies.

The Solution

To mitigate these financial vulnerabilities, an automated monitoring and detection solution hosted on the web application system was developed. The system breaks down traditional data silos and applies 21 distinct risk factors to replace manual checks with a predictive machine learning model that ranks transactions by risk, and enables cross-stream duplicate detection. Built entirely with open-source tools and existing resources.

The Impact

The solution provides a system that flags potential fraudulent transactions and prevents financial loss to the City of Edmonton. It enhances financial controls, detects anomalies in financial transactions and ensures compliance with regulations, thus minimizing the risk of fraud.

Lessons Learned

A key lesson learned relates to the benefit of clearly defining who or what will change operationally at the onset of the project. This has since become a project requirement and is baked into the team's intake process.

Launch year: 2021