|
| Estimation parameters | Description | Training data | Validation data |
| Model | % | Model | % |
|
| Accuracy | Accuracy of classification | PNN/GRNN | 80.00% | PNN/GRNN
TreeBoost | 73.76%
72.68% |
|
| True positive (TP) | Patients who are predicted as malignant among the malignant patients | PNN/GRNN | 39.76% | PNN/GRNN
MLP | 35.51%
33.07% |
|
| True negative (TN) | Patients who are predicted as nonmalignant among nonmalignant patients | PNN/GRNN | 46.34% | Decision tree
PNN/GRNN | 43.02%
41.88% |
|
| False positive (FP) | Patients who are predicted as malignant among nonmalignant patients | PNN/GRNN | 3.58% | PNN/GRNN | 12.83% |
|
| False negative (FN) | Patients who are predicted as nonmalignant among malignant patients | PNN/GRNN | 3.58% | PNN/GRNN
MLP | 4.41%
7.22% |
|
| Sensitivity | Probability to correctly predict malignancy | PNN/GRNN | 92.78% | PNN/GRNN
MLP | 87.67%
82.08% |
|
| Specificity | Probability to correctly predict nonmalignant cases | PNN/GRNN | 79.85% | Decision tree
PNN/GRNN | 72.06%
69.46% |
|
| Geometric mean of sensitivity and specificity | Geometric mean of sensitivity and specificity | PNN/GRNN | 80.55% | PNN/GRNN
TreeBoost | 74.05%
73.55% |
|
| Positive predictive value (PPV) | Proportion of patients with the disease who are correctly predicted to have the disease | PNN/GRNN | 71.49% | PNN/GRNN
TreeBoost | 62.86%
62.86% |
|
| Negative predictive value (NPV) | Proportion of patients who do not have the disease and who are correctly predicted as not having the disease | PNN/GRNN | 90.79% | PNN/GRNN
MLP | 88.17%
83.56% |
|
| Geometric mean of PPV and NPV | Geometric mean of PPV and NPV | PNN/GRNN | 79.14% | PNN/GRNN
TreeBoost | 72.23%
72.23% |
|
| Average gain for survival = | The gain shows how much of an improvement is provided by the model | PNN/GRNN | 1.40% | PNN/GRNN | 1.32% |
|
| Average gain for survival = | The gain shows how much of an improvement is provided by the model | PNN/GRNN | 1.65% | PNN/GRNN | 1.48% |
|
| Precision | Proportion of cases selected by the model that have the true value; precision is equal to PPV | PNN/GRNN | 71.49% | TreeBoost
PNN/GRNN | 62.86%
63.53% |
|
| Recall | Proportion of the true cases that are identified by the model; recall is equal to sensitivity | PNN/GRNN
GMDH | 91.8%
91.04% | PNN/GRNN
MLP | 86.67%
82.08% |
|
| -measure | It combines precision and recall to give an overall measure of the quality of the prediction | PNN/GRNN | 0.7715 | TreeBoost
PNN/GRNN | 0.7021
0.6593 |
|
| Area under ROC curve | Area under the Receive Operating Characteristic (ROC) curve for the model | PNN/GRNN | 0.892 | Decision Tree
PNN/GRNN | 0.835
0.821 |
|
