This paper examines the use of a linear model in combination with a multi-objective optimisation. A simple linear model is constructed and trained using data that has been automatically transformed based on skewness. These transformations, and their inverse, can then be used on the test data without having to make any assumptions of the underlying distribution of this data. Using nsga2 , the coefficients of the linear model are optimised across a pareto front using 3 objective functions, represent- ing 3 different error measurements. Although nsga2 produces a variety of non-dominated models across the pareto front, we show that the use of these models for creating an ensemble is inappropriate. Our main conclu- sion is that the use of pareto modelling for creating ensemble methods does not appear to be valuable, although there is some information that can be gained from examining the change in coefficient values of a linear model across the pareto front