Table 1.

The classification of accuracy, precision, recall, and F1 performance metrics when applying the new algorithm on 50% of the available data (n = 291/583) while exploring 20 combinations of hyperparameter values; 95% confidence intervals are shown between parentheses.

Merge (mins)	Logical operator	Edge default	Accuracy	Precision	Recall	F1
5	AND	Non-wear time	1.0 (± 0.0003)	0.9995 (± 0.0019)	1.0 (± 0.0)	0.9997 (± 0.0013)
4	AND	Non-wear time	0.9997 (± 0.0013)	0.9995 (± 0.0019)	0.9889 (± 0.0085)	0.9941 (± 0.0062)
4	OR	Wear time	0.999 (± 0.0026)	0.9551 (± 0.0168)	0.9993 (± 0.0022)	0.9767 (± 0.0123)
3	OR	Wear time	0.9989 (± 0.0026)	0.9552 (± 0.0168)	0.9979 (± 0.0038)	0.9761 (± 0.0124)
5	OR	Wear time	0.9989 (± 0.0027)	0.9498 (± 0.0177)	1.0 (± 0.0)	0.9742 (± 0.0129)
3	AND	Non-wear time	0.9974 (± 0.0041)	1.0 (± 0.0)	0.8785 (± 0.0265)	0.9353 (± 0.02)
2	OR	Wear time	0.9971 (± 0.0044)	0.9437 (± 0.0187)	0.9203 (± 0.022)	0.9319 (± 0.0205)
1	OR	Wear time	0.9962 (± 0.005)	0.9158 (± 0.0225)	0.9037 (± 0.0239)	0.9097 (± 0.0233)
3	OR	Non-wear time	0.9955 (± 0.0054)	0.828 (± 0.0306)	0.9979 (± 0.0038)	0.905 (± 0.0238)
2	OR	Non-wear time	0.9954 (± 0.0055)	0.8244 (± 0.0309)	0.9971 (± 0.0044)	0.9026 (± 0.0241)
1	OR	Non-wear time	0.9949 (± 0.0058)	0.8184 (± 0.0313)	0.9805 (± 0.0112)	0.8922 (± 0.0252)
4	OR	Non-wear time	0.9948 (± 0.0059)	0.8045 (± 0.0322)	0.9993 (± 0.0022)	0.8913 (± 0.0253)
5	OR	Non-wear time	0.9944 (± 0.0061)	0.7924 (± 0.0329)	1.0 (± 0.0)	0.8841 (± 0.026)
2	AND	Non-wear time	0.9954 (± 0.0055)	1.0 (± 0.0)	0.7862 (± 0.0333)	0.8803 (± 0.0264)
1	AND	Non-wear time	0.994 (± 0.0063)	1.0 (± 0.0)	0.7202 (± 0.0364)	0.8373 (± 0.03)
5	AND	Wear time	0.9908 (± 0.0077)	0.9991 (± 0.0025)	0.5736 (± 0.0401)	0.7288 (± 0.0361)
4	AND	Wear time	0.9906 (± 0.0078)	0.9991 (± 0.0025)	0.5625 (± 0.0403)	0.7197 (± 0.0365)
3	AND	Wear time	0.9888 (± 0.0086)	1.0 (± 0.0)	0.4763 (± 0.0405)	0.6452 (± 0.0388)
2	AND	Wear time	0.9887 (± 0.0086)	1.0 (± 0.0)	0.4742 (± 0.0405)	0.6433 (± 0.0389)
1	AND	Wear time	0.9873 (± 0.0091)	1.0 (± 0.0)	0.4082 (± 0.0399)	0.5797 (± 0.0401)

Merge (mins) = the merging of neighbouring candidate non-wear episodes to handle artificial movement. Logical operator = AND if both start and stop segments or OR if only one side of a candidate non-wear episode needs to be classified as true non-wear time to subsequently classify the candidate non-wear episode as an episode of true non-wear time. Edge default = the default classification of a candidate non-wear episode that has no start or end segment, such cases that occur right at the beginning or end of the acceleration data and default to wear or non-wear time.