Table 3.
Ranking of the top five features for the selected RF model across all temporal scales
| Days | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 14 | 30 | 90 | 365 | ||||||||||||
| F | I | F | I | F | I | F | I | F | I | F | I | F | I | F | I | F | I | F | I | F | I | |
| Ranking | ||||||||||||||||||||||
| 1 | EV | 15 | EV | 15 | EV | 15 | EV | 15 | TX | 16 | TX | 14 | EV | 14 | TX | 14 | TX | 15 | RH | 17 | RH | 16 |
| 2 | TX | 13 | TX | 14 | TX | 14 | TX | 14 | EV | 13 | EV | 13 | TX | 13 | TN | 13 | PR | 14 | EV | 15 | TX | 15 |
| 3 | TN | 11 | TN | 12 | PR | 13 | PR | 11 | TN | 12 | P | 13 | TN | 12 | EV | 12 | RH | 12 | PR | 14 | PR | 13 |
| 4 | RH | 11 | RH | 11 | TN | 12 | TN | 11 | PR | 11 | RH | 12 | RH | 12 | PR | 12 | TN | 12 | SD | 9 | EV | 13 |
| 5 | PR | 10 | PR | 9 | RH | 10 | RH | 10 | RH | 10 | TN | 9 | PR | 11 | RH | 11 | EV | 11 | TN | 9 | TN | 12 |
Each feature is accompanied by its importance, which has been calculated as the mean of 10 runs. Features involving atmospheric water levels (i.e. evapotranspiration and relative humidity) are found to be relevant to model tick activity in all temporal scales. These results are consistent with the ones provided by the general model. Interestingly, evapotranspiration is marked as the most relevant feature in very short-term time scales, whereas relative humidity is a better predictor for long time scales