Table 2.
Testing analysis (%) of the proposed IoT-enabled deep ensemble model for automated diagnosis of COVID-19 suspected cases on the four-class chest CT dataset when training to testing ratio is 65 : 35.
| Model | Accuracy | F-measure | AUC | Recall | Precision |
|---|---|---|---|---|---|
| JLM [16] | 97.46 | 97.84 | 97.65 | 97.98 | 98.08 |
| WSDL [17] | 96.96 | 97.17 | 97.06 | 97.46 | 97.88 |
| IPCNN [18] | 97.95 | 97.55 | 97.75 | 98.03 | 98.03 |
| DeCNN [19] | 97.36 | 97.36 | 97.18 | 97.27 | 98.11 |
| DLCRD [20] | 98.13 | 98.12 | 98.14 | 98.19 | 97.63 |
| PARL [22] | 97.11 | 97.52 | 97.31 | 97.78 | 97.93 |
| AGGDF [24] | 97.73 | 97.44 | 97.58 | 97.75 | 97.47 |
| GCNN [25] | 97.46 | 97.75 | 97.65 | 97.92 | 97.99 |
| GoogLeNet [53] | 97.77 | 97.33 | 97.55 | 97.69 | 97.94 |
| ResNet152V2 [44] | 96.96 | 97.89 | 97.42 | 97.74 | 97.67 |
| DenseNet201 [34] | 97.44 | 97.59 | 97.51 | 97.87 | 98.11 |
| IRNV2 [3] | 98.06 | 97.98 | 98.02 | 98.27 | 98.36 |
| Proposed | 98.97 | 98.75 | 98.57 | 98.58 | 98.56 |