Table 2 Performance of four machine learning models in predicting cognitive and sleep disorders.
AUC (95% CI) | Accuracy | F1 score | Sensitivity | Specificity | Precision | |
|---|---|---|---|---|---|---|
GP for predicting cognitive impairment (MoCA < 26 vs. MoCA ≥ 26) | ||||||
Training | 0.949 (0.900–0.998) | 0.877 | 0.905 | 0.864 | 0.905 | 0.950 |
Testing | 0.818 (0.610–1.000) | 0.706 | 0.762 | 0.727 | 0.667 | 0.800 |
DT for predicting subjective sleep quality (PSQI > 5 vs. PSQI ≤ 5) | ||||||
Training | 0.865 (0.770–0.959) | 0.846 | 0.886 | 0.907 | 0.727 | 0.867 |
Testing | 0.826 (0.616–1.000) | 0.824 | 0.870 | 0.909 | 0.667 | 0.833 |
GP for predicting subjective sleep quality (ISI > 7 vs. ISI ≤ 7) | ||||||
Training | 0.947 (0.888–1.000) | 0.862 | 0.830 | 0.759 | 0.944 | 0.917 |
Testing | 0.757 (0.492–1.000) | 0.765 | 0.714 | 0.713 | 0.800 | 0.714 |
DT for predicting excessive daytime sleepiness symptoms (ESS > 6 vs. ESS ≤ 6) | ||||||
Training | 0.923 (0.867–0.978) | 0.877 | 0.857 | 0.750 | 1.000 | 1.000 |
Testing | 0.875 (0.718–1.000) | 0.824 | 0.800 | 0.750 | 0.889 | 0.857 |
