Research Article
Usage of Probabilistic and General Regression Neural Network for Early Detection and Prevention of Oral Cancer
Table 6
Comparison of performance of classification models for training data.
| | Estimation parameters | Linear
regression | Decision
tree | TreeBoost | MLP | CCNN | PNN/GRNN |
| | Accuracy | 60.10% | 76.68% | 74.76% | 70.05% | 72.10% | 80.00% | | True positive (TP) | 1.37% | 30.44% | 32.80% | 31.02% | 33.46% | 39.76% | | True negative (TN) | 68.73% | 46.24% | 41.95% | 39.02% | 38.63% | 46.34% | | False positive (FP) | 0.98% | 13.46% | 17.80% | 20.68% | 21.07% | 6.46% | | False negative (FN) | 38.93% | 9.85% | 7.44% | 9.27% | 6.83% | 3.58% | | Sensitivity | 3.39% | 75.54% | 81.52% | 77.00% | 83.05% | 92.78% | | Specificity | 98.37% | 77.45% | 70.20% | 65.36% | 64.71% | 79.85% | | Geometric mean of sensitivity and specificity | 18.26% | 76.49% | 75.65% | 70.94% | 73.31% | 80.55% | | Positive predictive value (PPV) | 58.33% | 69.33% | 64.82% | 60.00% | 61.36% | 71.49% | | Negative predictive value (NPV) | 60.14% | 82.43% | 84.94% | 80.81% | 84.98% | 90.79% | | Geometric mean of PPV and NPV | 59.23% | 75.60% | 74.20% | 69.63% | 72.21% | 79.14% | | Average gain for survival = | 1.25% | 1.26% | 1.369% | 1.28% | 1.31% | 1.40% | | Average gain for survival = | 1.34% | 1.35% | 1.57% | 1.36% | 1.43% | 1.65% | | Precision | 58.33% | 69.33% | 64.82% | 60.00% | 61.36% | 71.49% | | Recall | 3.39% | 75.54% | 81.52% | 77.00% | 83.05% | 91.8% | | -measure | 0.0641 | 0.7231 | 0.7221 | 0.6744 | 0.7058 | 0.7715 | | Area under ROC curve | 0.722 | 0.835 | 0.8476 | 0.769 | 0.779 | 0.892 |
|
|
