Figure 2.

Flowchart for the algorithm used to construct the ANN architecture. Initially, the variables used to create the various ANN configurations were included. The program randomly selects values for each variable and generates an ANN, which is trained using the database; the data are divided into training, validation and test sets in accordance with the initial program selection. This process is repeated; the program compares the error for a network with previous networks, and the ANN with the lowest error is saved. When it encounters a stop variable, the program ends the cycle and displays the best result.