Symbolic Regression Model Comparison Approach Using Transmitted Variation

Abstract

Model evaluation in symbolic regression generated byGP is of critical importance for successful industrialapplications. Typically this model evaluation isachieved by a tradeoff between model complexity and Rsquared. The chapter introduces a model comparisonapproach based on the transmission of variation fromthe inputs to the output. The approach is illustratedwith three different data sets from real industrialapplications.