Jacob Burnim discusses Sift’s approach to building a ML system to detect fraud and abuse, including training models, handling imbalanced classes, sharing learning, measuring performance, etc..

Read full news article on InfoQ