Table 4.

Performance metric summaries comparing combined tuned and untuned model performance on PDBP validation dataset.

Data Modality	Combined	Combined; Untuned	Combined; Tuned
Stage	Untuned in PPMI as reference	Validation in PDBP	Validation in PDBP
Algorithm	AdaBoostClassifier	AdaBoostClassifier	AdaBoostClassifier
AUC (%)	89.72	83.84	85.03
Accuracy (%)	85.56	75.81	75.00
Balanced accuracy (%)	82.41	69.31	68.09
Log Loss	0.63	0.64	0.67
Sensitivity	0.89	0.93	0.93
Specificity	0.76	0.46	0.43
PPV	0.91	0.75	0.74
NPV	0.73	0.78	0.78