Fig. 3. Individual Test Prediction Accuracy (any-rater criterion).

The leave-one-subject-out cross-validation (LOSO-CV) classification accuracy of individual subtest smartphone-based prediction of the blinded MDS-UPDRS III subitems. Baselines and models are described in the legend to Fig. 2. Subtests relate to the corresponding MDS-UPDRS III subitems. Note that the Finger tapping MDS-UPDRS III subitem is repeated because it is predicted by two smartphone tasks (one or two target tapping).