Table 4.
Cities prediction statistics for Random Forest for CSD17 sampling date using proGenomes database with the top 16 features from RFE.
| CSD17 | |||
|---|---|---|---|
| Sensitivity | Specificity | Accuracy | |
| Kuala Lumpur | 1.00 | 1.00 | 1.00 |
| London | 1.00 | 0.992 | 0.996 |
| New York | 0.90 | 1.00 | 0.95 |
| Sendai | 0.92 | 0.98 | 0.95 |
| Stockholm | 0.90 | 0.96 | 0.93 |
| Seoul | 0.84 | 0.98 | 0.91 |
| San Francisco | 0.89 | 0.99 | 0.90 |
| Ilorin | 0.79 | 0.99 | 0.89 |
| Taipei | 0.80 | 0.97 | 0.88 |
| Singapore | 0.68 | 0.96 | 0.82 |
| Tokyo | 0.60 | 0.98 | 0.79 |
| Denver | 0.50 | 0.99 | 0.74 |
| Hong Kong | 0.53 | 0.95 | 0.74 |
| Zurich | 0.46 | 0.99 | 0.72 |
| Kiev | 0.42 | 0.98 | 0.70 |
| Offa | 0.40 | 0.996 | 0.70 |
| São Paulo | 0.36 | 0.99 | 0.68 |
| Doha | 0.33 | 1.00 | 0.67 |
The overall accuracy is 0.71 (0.65, 0.76). The data are sorted by balanced accuracy which is the average between sensitivity and specificity. For each method the tuning parameters are selected such that the best accuracy is achieved.