Research Article
Usage of Probabilistic and General Regression Neural Network for Early Detection and Prevention of Oral Cancer
Table 7
Comparison of performance of classification models for validation data.
| | Estimation parameters | Linear
regression | Decision
tree | Decision
tree forest | TreeBoost | MLP | CCNN | PNN/GRNN |
| | Accuracy | 61.27% | 68.88% | 67.41% | 72.68% | 69.76% | 68.29% | 73.76% | | True positive (TP) | 20.68% | 25.85% | 30.93% | 32.30% | 33.07% | 30.34% | 35.31% | | True negative (TN) | 40.59% | 43.02% | 36.49% | 40.49% | 36.68% | 37.95% | 41.88% | | False positive (FP) | 19.12% | 16.68% | 23.22% | 19.02% | 23.02% | 21.79% | 12.83% | | False negative (FN) | 19.61% | 14.44% | 9.37% | 8.29% | 7.22% | 9.95% | 4.41% | | Sensitivity | 51.33% | 64.16% | 76.76% | 79.52% | 82.08% | 75.30% | 87.67% | | Specificity | 67.97% | 72.06% | 61.11% | 68.03% | 61.44% | 63.56% | 69.46% | | Geometric mean of sensitivity and specificity | 59.07% | 68.00% | 68.49% | 73.55% | 71.01% | 69.18% | 74.05% | | Positive predictive value (PPV) | 51.96% | 60.78% | 57.12% | 62.86% | 58.96% | 58.24% | 62.86% | | Negative predictive value (NPV) | 67.42% | 74.87% | 79.57% | 83.00% | 83.56% | 79.23% | 88.17% | | Geometric mean of PPV and NPV | 59.19% | 67.46% | 67.42% | 72.23% | 70.19% | 67.93% | 72.23% | | Average gain for survival = | 1.149 | 1.15 | 1.273 | 1.274 | 1.28 | 1.26% | 1.32% | | Average gain for survival = | 1.17 | 1.17 | 1.324 | 1.413 | 1.31 | 1.32% | 1.48% | | Precision | 51.96% | 60.78% | 57.12% | 62.86% | 58.96% | 58.24% | 63.53% | | Recall | 51.33% | 64.16% | 76.76% | 79.52% | 82.08% | 75.30% | 86.67% | | -measure | 0.5164 | 0.6243 | 0.655 | 0.7021 | 0.6862 | 0.6568 | 0.6593 | | Area under ROC curve | 0.631 | 0.835 | 0.765 | 0.7705 | 0.739 | 0.731 | 0.821 |
|
|
