DailyLife

5-fold CV results

Binary classification (low vs. high) with 2,181 1-min segments of ['gsr', 'bpm', 'rri', 'temp'].

Arousal

$ python baseline.py -r ~/data/dailyLife2/datasets/60s -e ~/data/dailyLife2/metadata/esms_activity.csv -s 1 -t arousal --cv 'kfold' --splits 5 --shuffle --gpu

Metric	Random	Majority	Class ratio	Gaussian NB	XGBoost
acc.	0.501611	0.52132	0.499783	0.553404	0.55662
auroc	0.5	0.5	0.499221	0.577678	0.586021
bacc.	0.501703	0.5	0.499221	0.557775	0.555669
f1	0.510292	0.685352	0.515729	0.513229	0.575862

Valence

$ python baseline.py -r ~/data/dailyLife2/datasets/60s -e ~/data/dailyLife2/metadata/esms_activity.csv -s 1 -t valence --cv 'kfold' --splits 5 --shuffle --gpu

Metric	Random	Majority	Class ratio	Gaussian NB	XGBoost
acc.	0.491061	0.786795	0.661163	0.746965	0.75929
auroc	0.5	0.5	0.498554	0.521622	0.568744
bacc.	0.48922	0.5	0.498554	0.510779	0.521717
f1	0.603419	0.880677	0.783953	0.847594	0.859448