Using LLMs to detect fraud in credit card transactions

I am reading a lot of articles on the web that GenAI can be used to detect fraud in credit card transactions. But from my understanding of GenAI, it is good at dealing with unstructured data where as the credit card transactions data is structured data. I understand that one can build machine learning algorithms (other than GenAI) to do the task of fraud detection. Anyone please clarify and provide more info how GenAI can do this? If at all this is possible? How does GenAI fare against the traditional machine learning approach. Thanks.

I think it’s a very loose association.

Essentially, the generative AI is used to create a model of typical transactions.

Then the output of this model is compared with a new transaction, to see whether it is a statistical anomaly.

2 Likes

I agree with that. The association is weak. Instead use the traditional methods of anomaly detection. GenAI is usually for ‘generating stuff’ not for analyzing data. You generate text, images, audio, etc. ChatGPT can do that but it even struggles with small data in simple Excel formats. GenAI is not the best tool for this job-analysis. Use ML.