Table 2.

Confusion matrix of instantaneous predictions using greedy Gaussian segmentation from the 2 test experiments in the BREATHE dataset.

True categories	Xgboosta Predicted categories						Recall (%)	Precision (%)
	Lb	Rc	STd	STRe	STDf	WKg
L	38874	166	6920	830	3938	0	76.63	68.76
R	1587	31593	0	12402	791	11693	54.41	82.54
S	12483	0	38596	864	8030	154	64.19	72.65
STR	559	6505	1929	46751	2320	6156	72.80	71.17
STD	887	0	5127	0	54300	77	89.91	72.05
WK	2146	12	555	4846	5976	52455	79.49	74.37

^aXgboost specification: base_score=0.5, booster=“gbtree,” colsample_bylevel=1, colsample_bytree=1, gamma=0, learning_rate=0.1, max_delta_step=0, max_depth=3, min_child_weight=1, missing=None, n_estimators=200, n_jobs=1, nthread=None, objective=“multi:softprob,” random_state=0, reg_alpha=0, reg_lambda=1, scale_pos_weight=1, seed=None, silent=True, subsample=1. Overall accuracy: 79.4%.

^bL: lie.

^cR: run.

^dS: sit.

^eSTR: stair.

^fSTD: stand.

^gWK: walk.