a MLP model was constructed for the olive oil content data and
ber of hidden neurons was varied to optimise the performance.
25(a) shows the variable importance ranking of the MLP model
vip package. It ranked the fruit weight variable as the most
t one with the positive contribution to oil content as three
ed linear regression models. It also ranked the paste water
as a significantly negative correlated variable. Figure 4.25(b)
e fitness measurements of this MLP model. The R-square was
e F-statistic p value of the model was 1.33e−26. This improved
as due to the use of the MLP, which can explore nonlinearity
e data set.
(a) (b)
a) The vip visualisation of the variable ranking result of the MLP model for
oil content data. The model employs ten hidden neurons. (b) The fitness
nts for the MLP model.
e 4.26(a) shows the ranking result of the SVM models constructed
live oil content data. The ranking was implemented using the
ance package of the rminer package. All variables were given
rankings. In this model, the variable fruit width was ranked the
fruit weight was ranked the third while the paste water was ranked
d. Figure 4.26(b) shows the fitness measurements. The fitness of
M model was also significant. The R-square of the model was
d the F-statistic p value of the model was 6.05e−23.