TABLE A1.
Results of the random forest classifier evaluation, both for training data and when the classifier predicted to test data.
| Metric | Calculation | Mean (training) | SD (training) | Mean (test) | SD (test) |
|---|---|---|---|---|---|
| Misclassification rate | 1 − (TP + TN/total calls) | 0.111 | 0.00023 | 0.107 | 0.00052 |
| False negative rate | FN/FN + TP | 0.100 | 0.00026 | 0.098 | 0.00063 |
| True positive rate | TP/TP + FN | 0.899 | 0.00026 | 0.901 | 0.00063 |
| False positive rate | FP/FP + TN | 0.126 | 0.00041 | 0.121 | 0.00095 |
| True negative rate | TN/TN + FP | 0.873 | 0.00041 | 0.878 | 0.00095 |
Note: Each bootstrap iteration trained the model on a randomly selected 75% of the data, and predicted to the withheld 25%, and this process was repeated over 100 runs.