Table 2.
Objective 1: results of the application of the MOGA algorithm on each of the six different feature datasets
| Case | Number of Features | Writing Pattern | ANN Topology | Accuracy | |
|---|---|---|---|---|---|
| Number of Neurons | Activation Function | ||||
| 1 | 41 | 1 | 186/15/2 | logsig/logsig/softmax | 90.76% |
| 2 | 43 | 2 | 44/10/2 | logsig/logsig/softmax | 92.98% |
| 3 | 43 | 3 | 232/82/7/2 | logsig/logsig/logsig/softmax | 95.95% |
| 4 | 6 | 1 | 222/25/2 | logsig/logsig/softmax | 93.78% |
| 5 | 6 | 2 | 246/12/2 | logsig/logsig/softmax | 91.58% |
| 6 | 7 | 3 | 45/114/21/2 | satlins/tansig/logsig/softmax | 96.85% |
The output layer configuration was preliminarily fixed with two neurons and softmax as activation function