Nonlinear nonnegative multiregressions based on Choquet integrals

Using a nonadditive set function to describe the interaction among attribut es, a new nonlinear nonnegative multiregression is established based on Cho quet integrals with respect to the set function. Regarding the values of th e set function as unknown regression parameters, an evolutionary computatio n can be used to determine them when necessary data are available. Such a m odel is a generalization of the traditional linear multiregression. It prov ides an effective regression tool in some real problems where the linear mu ltiregression model and the second-order multiregression model fail. This n ew method has a wide applicability in the areas of information fusion and d ata mining, as well as in the areas of decision making, image processing, p attern recognition, medical and industrial diagnoses, and expert systems. ( C) 2000 Elsevier Science Inc. All rights reserved.