On Combining Classification Algorithms (Paperback)

The ability of a chosen classification algorithm to induce a good generalization depends on how appropriate its representation language used to express generalizations of the examples is for the given task. Since different learning algorithms employ different knowledge representations and search heuristics, different search spaces are explored and diverse results are obtained. The problem of finding the appropriate model for a given task is an active research area. In this dissertation, instead of looking for methods that fit the data using a single representation language, we present a family of algorithms, under the generic name of {em Cascade Generalization}, whose search spaces contains models that use different representation languages.