Using An Evolutionary Strategy to Select Input Features for a Neural Network Classifier

Modelling in high dimensional spaces is usually following a feature selection process. Feature selection and feature creation are two of the most important and difficult tasks in the field of pattern recognition. It involves determination of a good feature subset, given a set of feature candidates. The present method is one approach to improve the pattern classification performance using a feature selection process in conjunction with a neural network classifier an implied feed forward architecture. The feature selection part is based on an evolutionary strategy and the subsequent classification step. Both have been combined.