**3.7. Residual Analyses for the genetic individual with the best AIC**

At the end of all the iterations, if no genetic individual is approved in the formal evaluations, the GP algorithm will select the solution with the best AIC for residual analysis. The residual analysis allows the evaluation of the assumptions about a model [12].

So, in this work, the residual analysis is divided in two stages:

	- Diagram of distribution of accumulated errors, to quantify the distance between the estimates given by the LRM and the data of the training set;
	- Q-Q Plots and Histograms, to check the assumptions about the error probability distributions;
	- Diagram of residuals dispersion against the fitted values, the check the assumption of homoscedasticity;
	- Diagram of dispersion of the residuals, to check the absence of autocorrelation among the errors.
