Microsoft Uses Machine Learning and Optimization to Reduce E-Commerce Fraud
Jay Nanduri, Yuting Jia, Anand Oka, John Beaver, Yung-Wen Liu
Abstract
The authors discuss Microsoft’s development of a fraud-management system that uses customized long-term and short-term sequential machine learning models to detect both historical and emerging fraud patterns. It also makes rapid real-time optimal decisions using a dynamic programming approach to optimize long-term profit by taking into account decisions made by multiple parties (e.g., banks issuing credit cards).
Topics & Concepts
Computer scienceTerm (time)Profit (economics)Dynamic programmingComputer securityArtificial intelligenceEconomicsAlgorithmMicroeconomicsQuantum mechanicsPhysicsImbalanced Data Classification TechniquesStock Market Forecasting MethodsData Stream Mining Techniques