Abstract:
The purpose of this paper is to illustrate methods which can be used to select relevant input variables for non-linear regression. Relevant variables are ones which are though to explain in some extent the behaviour of the process we want model. We propose a method based on the self-organised map which allows to compute the local linear correlation between variables in neighbour subspaces. The graphical ordered plot of the obtained local correlations allows to study the non-linear dependencies of variables and suggests relevant ones.