Harnessing Machine Learning to Combat E-commerce Fraud
summary
Claims concrete numbers and engineering choices: rule systems show ~28% false positives and $3.6B lost to false declines, while ML reports 96% detection accuracy and an 81% drop in false positives. The stack uses supervised models (random forest, gradient boosting, neural nets) plus unsupervised tools (isolation forest, autoencoder, clustering), feature engineering (temporal, network, behavior), and a distributed fault tolerant microserbased architecture that makes decisions in 70–150 ms at million‑transaction scale; hybrid + manual‑review loops cut review time ~30% and raise accuracy ~40%.
AI-written summary · about
chapters (13)
- 0:00Introduction to E-commerce Fraud
- 0:47The Growing Threat of E-commerce Fraud
- 1:38Limitations of Traditional Rule-Based Systems
- 2:20The Power of Machine Learning in Fraud Detection
- 3:01Supervised Learning Techniques
- 4:34Unsupervised Learning Techniques
- 5:37Feature Engineering for Fraud Detection
- 7:11Scaling Fraud Detection Systems
- 8:16Hybrid Systems: Combining Rules and ML
- 9:16The Role of Manual Reviews
- 10:34Business Value of ML in Fraud Prevention
- 11:18Future Trends in Fraud Detection
- 12:53Conclusion
related talks
45mAnti-Fraud in Reverse and the Use of ML Methods in It
16mLeveraging ML to Create a Reliable Anti-Fraud System
26mAI-Powered Fraud Detection in Cloud-Native Digital Payments
37mThe Role of Machine Learning in Today’s Payment Ecosystems
39mFraud Detection: Using ML to Identify and Manage Fraudulent Activities
41mUsing Machine Learning to Fight Fraud
12mHarnessing AI & LLMs for Next-Gen Digital Payment Security
40mFraud detection without labeled data