Nonlinear continuum regression: an evolutionary
approach
Transactions of the Institute of Measurement and
Control, 22 (2):
125--140 (2000)DOI: doi:10.1177/014233120002200202
Abstract
genetic programming is combined with continuum
regression to produce two novel non-linear continuum
regression algorithms. The first is a sequential
algorithm while the second adopts a team-based
strategy. Having discussed continuum regression, the
modifications required to extend the algorithm for
non-linear modelling are outlined. The results of two
case studies are then presented: the development of an
inferential model of a food extrusion process and an
input-output model of an industrial bioreactor. The
superior performance of the sequential continuum
regression algorithm, as compared to a similar
sequential nonlinear partial least squares algorithm,
is demonstrated. These applications clearly demonstrate
that the team-based continuum regression strategy
significantly outperforms both sequential approaches.