Table 2.
Comparison of the predictive performance of different models in optimal feature subset in validation set.
| Variable | LR (95% CI) | QDA (95% CI) | NB (95% CI) | SVM (95% CI) | AdaBoost (95% CI) | BP (95% CI) | p value |
|---|---|---|---|---|---|---|---|
| SEN | 65.4% (55.2-74.5%)a,c,d,e | 83.2% (75.7-90.6%)b | 81.2% (73.4-88.9%)b | 71.3% (62.3-80.3%) | 80.2% (72.3-88.1%)b | 83.2% (75.7-90.6%)b | 0.008 |
| SPE | 90.0% (85.4-93.6%) | 77.4% (71.5-82.4%) | 78.3% (72.4-83.2%) | 83.9% (78.6-88.3%) | 80.4% (75.3-85.6%) | 76.5% (71.0-82.0%) | 0.002 |
| FPR | 10.0% (6.4-14.9%) | 22.6% (17.6-28.5%) | 21.7% (16.8-27.6%) | 16.1% (11.7-21.4%) | 19.57% (14.4-24.7%) | 23.5% (18.0-29.0%) | 0.002 |
| FNR | 35.6% (25.5-46.8%)a,c,d,e | 16.8% (9.4-24.3%)b | 9.1% (11.1-26.6%)b | 28.7% (10.7-37.7%) | 19.8% (11.9-27.7%)b | 26.8% (9.4-24.3%)b | 0.008 |
| PPV | 73.3% (64.0-82.6%) | 61.3% (53.1-69.6%)a | 61.7% (53.3-70.0%)b,c | 65.5% (56.4-74.5%)a | 64.3% (55.8-72.8%) | 60.9% (52.6-69.1%) | 0.437 |
| NPV | 85.5% (81.0-90.0%) | 91.2% (87.2-95.3%) | 90.4% (86.3-94.5%) | 86.9% (82.4-91.4%) | 90.2% (86.1-94.3%) | 91.2% (87.2-95.2%) | 0.239 |
| Accuracy | 82.2% (78.0-86.3%) | 78.9% (74.4-83.3%) | 78.9% (74.4-83.3%) | 79.8% (75.4-81.4%) | 80.4% (76.1-84.7%) | 78.5% (74.1-83.0%) | 0.862 |
| AUC | 0.840 (0.796-0.878) | 0.865 (0.824-0.900) | 0.864 (0.823-0.899) | 0.839 (0.795-0.877) | 0.863 (0.821-0.898) | 0.862 (0.820-0.897) | / |
aCompared with QDA, p < 0.05; bCompared with LR, p < 0.05; cCompared with NB, p < 0.05; dCompared with AdaBoost, p < 0.05; eCompared with BP, p < 0.05. p value denoted the overall statistical result for the four models.