Estimation of Risk and Probability: A Machine Learning Approach

The book includes a chapter introducing readers to learning machines. Logistic regression is also explained briefly to make the connection with the presented analyses and to illustrate the assumptions required by logistic regression that are no longer needed in the learning machine context. Detail on how to run learning machines and how to calculate consistent probability estimates is also presented as well as additional topics of coverage including: prediction problems and the need to calculate probabilities; traditional approaches to analyzing prediction problems and computing probabilities; logistic regression and its limitations; the importance of consistency in computing probabilities; validation, and why results must be validated, and techniques for doing validation; and techniques for ranking and selecting predictors. Throughout the book, R and SAS software are used to carry out the discussed analyses and showcase the differences between the obtained results and those gathered from using logistic regression. End-of-chapter notes and appendices touch upon the mathematical theory, allowing the book to focus on worked examples and basic statistical concepts, such as the assumptions that logistic regression models require.