Fraud Detection - Banking - Brainstorming

Question

Morning guys!

My goal with this topic is to garner ideas and brainstorm about how to create a fraud detection analysis with transactions in a banking environment.

I have all data like client name, frequency of payment, different types of payments, banking account, TAX ID, address, etc etc.

Has anyone tried to make a similar workflow and can kindly share with me? Any ideas of what predictive and prescriptive tools should I use to make that happen?

Thanks!!!

IraWatt · Answer

No worries @Joker_Hazard ! by labelled I just meant that your data came with examples of what was and what was not fraud, this would let you train a model to use on new examples. There are a few algorithms to look at for unsupervised methods Anomaly Detection Algorithms: in Data Mining (With Comparison) (intellspot.com). The simplest KNN I think Alteryx has here: Find Nearest Neighbors Tool | Alteryx Help

Joker_Hazard · Answer

@IraWatt Thanks for comment!

Would you elucidate what is a "label"? If you may as well provide an example of unsupervised methods please!

Excuse my beginner doubt.

Thanks

IraWatt · Answer

Hey @Joker_Hazard,

This is quite a common machine learning problem. If you have data which contains labels to whether a row was fraud or not you can use Alteryx's logistic regression tool or one of Alteryx's many other ML tools to build a model which can identify and predict fraud. If you don't have labels there are unsupervised methods for anomaly detection which can be used to detect fraud.