CASE-BASED PREDICTIONS: AN AXIOMATIC APPROACH TO PREDICTION, CLASSIFICATION AND STATISTICAL LEARNING

The book presents an axiomatic approach to the problems of prediction, classification, and statistical learning. Using methodologies from axiomatic decision theory, and, in particular, the authors' case-based decision theory, the present studies attempt to ask what inductive conclusions can be derived from existing databases. It is shown that simple consistency rules lead to similarity-weighted aggregation, akin to kernel-based methods. It is suggested that the similarity function be estimated from the data. The incorporation of rule-based reasoning is discussed.