Sequential selection of discrete features for neural networks - A Bayesian approach to building a cascade

A feature selection procedure is used to successively remove features one-by-one from a statistical classi®er by an iterative backward search. Each classi®er uses a smaller subset of features than the classi®er in the previous iteration. The classi®ers are subsequently combined into a cascade. Each classi®er in the cascade should classify cases to which a reliable class label can be assigned. Other cases should be propagated to the next classi®er which uses also the value of a new feature. Experiments demonstrate the feasibility of building cascades of classi®ers (neural networks for prediction of atrial ®brillation (FA)) using a backward search scheme for feature selection. Ó 1999 Elsevier Science B.V. All rights reserved.