Construction and Assessment of Classification Rules

'Construction and Assessment of Classification Rules', David J. Hand, Published by John Wiley and Sons, Chichester, 1997. ISBN 0-471-96583-9. Supervised classification is a paradigmatic statistical problem: in its own microcosm it includes all of the major issues of statistics. The many methods which have been developed for tackling such problems are described here, from classical linear discriminant analysis through to recursive partitioning and neural networks. Particular emphasis is placed on evaluating such rules and assessing their quality. The common measures of misclassification rate and ROC curve analysis are rarely sufficient. Detailed discussion is given of several important practical applications and the relative merits and properties of the various methods are described.

Go to Statistics section home page

Go to Imperial College home page