AI powered Fraud detection system

James Longbottom

Based at reducing the risk of fraud in countries with high levels of online banking and transactional fraud. The system was based from a machine learning core that would be trained from a large amount of features that could classify fraud or high risk transactions. The system would uses NLP (Natural language processing) to evaluate text from emails and other communications that could point to other levels of risk.

5 min read

Share this post:

Overview of system

Training and learning system – a focused engine would be trained using test data. Once fully trained it could be then used to detect fraud. Once in use this learning system would be used to improve the system over time, the more it was used.

Dashboard – this is for each user to monitor the system and view any fraud, this dashboard can also create notifications to users to prevent fraud. Tagging of correct instances of fraud is also possible here.

Detailed report

Since this project was aimed at countries that experience high levels of fraud, a self learning AI was essential, it was designed to operate at scale, where multiple users could contribute to identify fraud and the signs of it, thus saving countless others from being exposed to the same patterns that fraud deploys on scale.