Input Selection

Abstract

In this chapter we describe a general method for reducing the number of potential inputs to a model. This is required because a number of modelling techniques, especially neural networks, can only use a limited number of inputs because of the parameterisation of the model and the limited number of data points available. We note that a number of techniques (e.g. adaptive lag and linear RVM) have their own selection techniques, and so avoid some of the limitations in the method described below. We also note that before we get to this stage we will have been through a data reduction stage using, for example, the PCA techniques described earlier.