Table 2.
Internal validation performance for the prediction of curative resection of undifferentiated type of early gastric cancer by using 18 machine learning classifiers.
| Machine learning classifier | Accuracy (%) (95% CI) | Precision (%) (95% CI) | Recall (%) (95% CI) | F1 score (%) (95% CI) | |
| Gaussian Naïve Bayes | 73.8 (68.6-79.0) | 86.2 (82.1-90.3) | 77.2 (72.2-82.2) | 81.5 (76.9-86.1) | |
| Linear discriminant analysis classifier | 76.4 (71.3-81.5) | 77.4 (72.4-82.4) | 96.5 (94.3-98.7) | 85.9 (81.8-90.0) | |
| Logistic regression classifier | 77.5 (72.5-82.5) | 80.5 (75.8-85.2) | 92.1 (88.9-95.3) | 85.9 (81.8-90.0) | |
| Linear support vector machine classifier | 74.5 (69.3-79.7) | 74.5 (69.3-79.7) | 99.9 (98.8-99.9) | 85.4 (81.2-89.6) | |
| Stochastic gradient descent classifier | 74.5 (69.3-79.7) | 77.6 (72.6-82.6) | 92.6 (89.5-95.7) | 84.4 (80.1-88.7) | |
| Decision tree classifier | 74.5 (69.3-79.7) | 74.5 (69.3-79.7) | 99.9 (98.8-99.9) | 85.4 (81.2-89.6) | |
| k-nearest neighbors classifier | 72.0 (66.7-77.3) | 78.1 (73.2-83.0) | 86.6 (82.5-90.7) | 82.2 (77.6-86.8) | |
| Deep neural network | 77.9 (73.0-82.8) | 80.6 (75.9-85.3) | 92.6 (89.5-95.7) | 86.2 (82.1-90.3) | |
| Ensemble (bagging) | |||||
|
|
Bagging classifier | 72.0 (66.7-77.3) | 81.2 (76.5-85.9) | 81.2 (76.5-85.9) | 81.2 (76.5-85.9) |
|
|
Random forest classifier | 72.7 (67.4-78.0) | 80.2 (75.5-84.9) | 84.2 (79.9-88.5) | 82.1 (77.5-86.7) |
|
|
Voting classifier | 84.5 (80.2-88.8) | 88.1 (84.2-92.0) | 91.6 (88.3-94.9) | 89.8 (86.2-93.4) |
| Ensemble (boosting) | |||||
|
|
Gradient boosting classifier | 77.5 (72.5-82.5) | 80.5 (75.8-85.2) | 92.1 (88.9-95.3) | 85.9 (81.8-90.0) |
|
|
Adaptive boosting classifier | 77.9 (73.0-82.8) | 81.1 (76.4-85.8) | 91.6 (88.3-94.9) | 86.0 (81.9-90.1) |
|
|
Categorical boosting classifier | 84.1 (79.7-88.5) | 83.8 (79.4-88.2) | 97.5 (95.6-99.4) | 90.2 (86.7-93.7) |
|
|
Extreme gradient boosting classifier | 93.4 (90.4-96.4) | 92.6 (89.5-95.7) | 99.0 (97.8-99.9) | 95.7 (93.3-98.1) |
|
|
Light gradient boosting machine classifier |
75.6 (70.6-80.8) | 80.9 (76.2-85.6) | 88.1 (84.2-92.0) | 84.4 (80.1-88.7) |
|
|
Histogram-based gradient boosting classifier |
85.2 (81.0-89.4) | 84.9 (80.689.2) | 97.5 (95.6-99.4) | 90.8 (87.4-94.2) |
| Ensemble (stacking) | 75.6 (70.5-80.7) | 78.6 (73.7-83.5) | 92.6 (89.5-95.7) | 85.0 (80.7-89.3) | |