Table 3 Results of the multimodal approach when the features are randomly divided. As in Table 2, missing continuous and categorical values are imputed by the mean and the mode, respectively, as reported in “Data preparation” section
From: Correction: Machine learning predicts pulmonary long Covid sequelae using clinical data
Meta-learner | Performance (%) | ||||
|---|---|---|---|---|---|
Accuracy | Sensitivity | Specificity | AUC | F1-score | |
Bayesian classifier Decision Tree SVM XGBoost | 86.4 ± 1.6 76.4 ± 1.4 91.6 ± 0.7 81.4 ± 0.9 | 69.2 ± 3.5 55.2 ± 2.9 79.2 ± 2.6 59.3 ± 2.1 | 95.8 ± 1.3 88.8 ± 1.1 99.2 ± 0.7 93.3 ± 1.2 | 92.7 ± 1.8 83.9 ± 2.1 98.0 ± 0.5 88.0 ± 1.7 | 78.3 ± 2.7 63.1 ± 2.5 86.5 ± 1.3 68.7 ± 1.6 |