Table 3 Performance of each of the three XGBoost classification models

From: Smartphone-based prediction of dopaminergic deficit in prodromal and manifest Parkinson’s disease

 

AUC ± SD (95% CI)

AUC adjusted ± SD (95% CI)

Sensitivity ± SD

Specificity ± SD

Balanced accuracy ± SD (Sensitivity + Specificity)/2

MDS-UPDRS-III

0.85 ± 0.04(0.72–0.90)

0.79 ± 0.05(0.68-0.88)

0.68 ± 0.23

0.89 ± 0.10

0.79 ± 0.13

Smartphone features (XGBoost)

0.84 ± 0.11 (0.74–0.90)

0.80 ± 0.05 (0.72–0.88)

0.68 ± 0.11

0.80 ± 0.19

0.74 ± 0.11

Smartphone features + MDS-UPDRS-III (XGBoost)

0.88 ± 0.05 (0.75-0.92)

0.82 ± 0.05 (0.72–0.90)

0.76 ± 0.10

0.91 ± 0.11

0.84 ± 0.07

  1. The AUC adjusted for the effect of sex is also reported (AUC adjusted). Balanced accuracy is the average of sensitivity and specificity. SD: standard deviation of the error metric.
  2. Reported using the Area-Under-the-Curve (AUC) with 95% confidence intervals (CI), sensitivity, and specificity.
Back to article page