Predicting cognitive scores from wearable-based digital physiological features using machine learning: data from a clinical trial in mild cognitive impairment

Table 4 Average models’ performance (Pearson r) in predicting NTB scores across varying feature selection criteria

Cross-validation method	Cognitive measure	Machine Learning algorithms
Cross-validation method	Cognitive measure	ElasticNet	Random Forest	XGBoost
Subject-based	Global Cognition	0.51	0.23	0.26
	Executive Function	0.67	0.20	0.31
	Processing Speed	0.43	0.08	0.09
	Memory Immediate	0.42	0.12	0.17
	Memory Delayed	0.42	0.16	0.21
Interval-based	Global Cognition	0.73	0.91	0.87
	Executive Function	0.79	0.89	0.83
	Processing Speed	0.66	0.83	0.80
	Memory Immediate	0.64	0.91	0.89
	Memory Delayed	0.67	0.84	0.82

ISSN: 1741-7015