The Elements of Statistical Learning

Machine learning and data mining are set within one statistical framework by Trevor Hastie, Robert Tibshirani and Jerome Friedman, with the emphasis on concepts rather than mathematics. Coverage runs from supervised learning, or prediction, to unsupervised learning, with topics such as boosting, support vector machines, neural networks and classification trees. The newer edition adds random forests, graphical models, ensemble methods, least angle regression and path algorithms for the lasso, spectral clustering and non-negative matrix factorization, plus a chapter on data with more variables than observations. Examples come from medicine, biology, finance and marketing, with many color graphics. It serves statisticians and anyone using data mining in science or industry.

Nassim Nicholas Taleb praised it as a very comprehensive and technical reference on the inner workings of machine learning, adding that he owns two copies.