Table 5.

Effect of window length for feature extraction on micro-averaged accuracy using random forest model for individual activity recognition, locomotion detection, sedentary activity detection, activity intensity level classification, and major body movement location detection. The best performance for each classification task is presented in italics.

Classification task: window length	Classification accuracy (seconds)
	1	2	4	8	15	16
Individual activity recognition	.45	.47	.56	.60	.64	.64
Locomotion	1	1	1	1	1	1
Sedentary	.96	.97	.97	.98	.98	.98
Intensity	.80	.81	.82	.84	.87	.87
Major body movement location	.78	.80	.81	.83	.85	.85