Table 4.
Classification performance results of the neural network with 9 and 4 predictors according to the number of hidden layers (HL) and the number of neurons in the HL*
| Models | Number of features | Accuracy | Sensitivity | Specificity | Fowlkes-Mallows index | Matthews correlation coefficient | |
|---|---|---|---|---|---|---|---|
|
M5-9
(HL-2)
(HL1-11)
(HL2-100) |
9 |
training |
80.5 |
75.0 |
84.8 |
77.2 |
60.2 |
| validation |
82.4 |
85.7 |
80.0 |
80.2 |
64.8 |
||
| test |
70.6 |
77.8 |
62.5 |
73.8 |
40.9 |
||
| all |
79.3 |
76.9 |
81.3 |
76.9. |
58.2 |
||
| M5-4 (HL-2) (HL1-11) (HL2-100) | 4 | training |
90.2 |
86.1 |
93.5 |
88.6 |
80.2 |
| validation |
88.2 |
100.0 |
80.0 |
88.2 |
78.9 |
||
| test |
82.4 |
77.8 |
87.5 |
82.5 |
65.3 |
||
| all | 88.8 | 86.5 | 90.6 | 87.4 | 77.3 |