Table 4 Performance results of the models developed in this research
Model | Optimal hyperparameters | Accuracy | Sensitivity | Specificity |
|---|---|---|---|---|
PSO-FCM | Initial population = 200 | 0.95 | 0.96 | 0.94 |
Activation function = sigmoid | ||||
Inference function = modified-Kosko | ||||
ANN | Hidden layer units = 256 | 0.99 | 0.99 | 0.99 |
Learning rate = 0.0001 | ||||
Activation function = ReLU | ||||
Optimizer = Adam | ||||
Learning type = constant | ||||
SVM | Kernel = radial | 0.99 | 1.00 | 0.99 |
C = 1.0 | ||||
gamma = 1.0 | ||||
XGBoost | Randomly Selected Predictors = 5 | 1.00 | 1.00 | 1.00 |
Trees = 1000 | ||||
Minimal node size = 2 | ||||
Tree depth = 10 | ||||
Learning rate = 0.08321 | ||||
Loss reduction = 7 x 10-9 | ||||
Proportion observations sampled = 0.8852 |