A methodology that seeks to enhance model prediction performance is presented. The method involves generating multiple auxiliary models that capture relationships between attributes as a function of each other. Such information serves to generate additional informative columns in the dataset that can potentially enhance target prediction. A proof of case and related code is provided.