Computer Systems That Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning and Expert Systems

This is a classic for anyone interested in machine learning, data mining or predictive statistics. Though it is beginning to age, it covers essential aspects of empirical modeling still not covered by many more recent titles (!). A subsequent effort by one of the authors, "Predictive Data Mining" is a bit more current though shorter on the fundamentals.

This book gives a good cohesive introduction to the basic algorithms from Statistics, Machine Learning and Pattern Recognition research. These include Nearest Neighbor, Decision Trees, Bayesian Networks, and Neural Networks. The main value of the book however is its coverage of techniques that 1) estimate a model's accuracy, and 2) select a 'good' model. This book offers the reader a solid foundation to what we are trying to achieve: to get at the objective truth.