3. Modeling Phase
Four sequential stages were followed to develop the auto-associate networks and traditional ANNs for seven databases. In the first
stage, the ANN architecture was determined based on problem characteristics, and then input and output categories were selected
accordingly. In this step the training, testing, and validation datasets are also determined. In the second stage, the network was trained
and tested on the datasets to obtain the optimum number of hidden nodes and iterations for the ANN architecture determined in stage
one. In the third stage, the best performing network obtained from the second stage was validated on the validation database. If accuracy
measures from training, testing and validation database are highly comparable, then the model may not be trained on all data.