**3.6. Formal evaluation of a linear regression model**

Once an iteration of the proposed GP algorithm is ended, the best solution found in the iteration is formally evaluated. In linear regression, assumptions about the fitted model must be considered so that the results can be reliable. So, the evaluation process consists in verifying, by residual inference, the assumptions of normality, homoscedasticity and independence about the distribution of errors of the fitted LRM. We used the following adherence tests:


If the result of any of these tests is not positive and the maximum number of iterations was not reached, the GP algorithm will start a new evolution iteration through the generation of a new starting population and will follow the flow presented in Figure 2. Otherwise, the algorithm presents the LRM as final solution.
