Volume 3

Volume 3 - Machine Learning Engineering

Volume 3 builds the classical models that still run most production prediction, from least squares and gradient-boosted trees to clustering and recommenders, and measures each one on real data: 442 diabetes patients, 569 breast-cancer tumours, 1,797 handwritten digits, the published ULB card-fraud figures and Banking77's 77 banking intents. Each teaching chapter pairs its method with the way it produces a plausible wrong number, and ends by choosing among real options with numbers and naming the evidence that would reverse the choice. The review chapter applies the whole volume to one fraud-scoring design.