*3.2. Test 2—Influence of the Amount of Input Data*

Test 2, where the amount of data introduced for training was varied, revealed that the ANN with the lowest ME was T2\_13, with an average error of −0.08 min. With an MAE of 4.85 min, T2\_15 was the ANN with the best result. The ANN with the lowest MSE was T2\_14 with 40.18 min. The smallest EMax recorded was 16.73 min, which occurred in T2\_09. The lowest PE belonged to T2\_15, like T1\_07, and with a PE of 2.52%, which ensured the best performance (please see Table 4).

From the training of the networks for Test 2, it was verified that:


The graph in Figure 5 shows the percentage error variation as a function of the number of input data. The error decreases as the number of parts increases. From 250 parts onwards, the percentage error is reduced to a lesser extent.


**Table 4.** Regression and error of the ANNs from Test 2.

**Figure 5.** Percentage error (PE) variation as a function of the amount of input data.
