3.2. Modelling and performance evaluation
Table 3 presents the performance metrics of the individual and ensemble
models for each COVID-19 severity category. Of the individual models,
the NN model outperformed SVM and QUEST algorithms based on the
performance metrics in the training and testing datasets. However,
ensemble approaches (i.e., HCWS, VS, CWVS) gave better predictions as
compared to the individual models regarding all the evaluation metrics.
As the ensemble models’ estimates were compared, VS achieved slightly
better prediction performance than the HCWS and CWVS algorithms.
Predictor importance values for each separate model are summarized in
Table 4. Based on the estimates of the best-performing individual model
(i.e., NN), the three most important predictors were age, Favipiravir
use, and the presence of dyspnea, respectively. The predictor
significance values of other individual models are also shown in Table
4.