Classifier models | Sensitivity (%) | Specificity (%) | Precision (%) | NPV (%) | BA (%) | F1-score |
---|
MEE proposed | 73.5 | 88.3 | 68.3 | 91.5 | 80.9 | 70.8 |
AdaBoost | 54.2 | 93.3 | 73.8 | 87.0 | 73.7 | 62.5 |
MLP | 76.3 | 78.4 | 54.4 | 91.6 | 77.4 | 63.5 |
NB | 48.5 | 90.0 | 61.1 | 85.4 | 69.4 | 54.1 |
DT | 59.1 | 87.0 | 66.0 | 87.5 | 73.1 | 62.4 |
KNN | 52.8 | 93.5 | 73.4 | 87.4 | 74.6 | 61.4 |
LR | 75.0 | 79.5 | 55.2 | 91.3 | 77.3 | 63.6 |
RF | 68.6 | 86.0 | 63.1 | 89.9 | 77.3 | 65.7 |
SVM | 65.6 | 89.0 | 66.9 | 89.4 | 77.3 | 66.2 |
- Sensitivity: ratio between the AD converter subjects correctly labeled by the algorithm and all subjects that actually converted; Specificity: ratio between the non-AD converter subjects correctly labeled by the algorithm and all subjects that have not actually converted; Precision: ratio between the correctly AD converter subjects labeled by the algorithm and the AD converters; Negative predictive value (NPV): the proportion of predicted negatives which are real negatives. It reflects the probability that a predicted negative is a true negative; Balanced accuracy (BA): the average between sensitivity and specificity; F1-score: the harmonic average of the sensitivity and precision