Task leaderboard

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery)

Last updated

Task leaderboard

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery)

3 reported results · Leaderboard page

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): Jang, Area under ROC. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationArea under ROC↑ Higher is better
1XGBoost
Evaluation details · caveat

Site: wrist

Caveat: Not subject-independent; 3,969 day points, 254 with panic; accuracy 0.919.

0.905 (95% CI 0.885–0.924)
2GradientBoost
Evaluation details · caveat

Site: wrist

Caveat: Same.

0.826
3Random forest
Evaluation details · caveat

Site: wrist

Caveat: Same.

0.730

Rows are ranked only within this task × dataset × metric and one protocol group. Protocols can still differ in detail; read each method’s evaluation details before comparing.

Recent changes

From the hub's dated sync history: new results as they are added, and SOTA stars as they are awarded or move.

All reported results for this task (6 benchmarks)

Rankings apply within each dataset and metric. Open evaluation details for differences in protocols.

same: accuracy (2-level, window CV)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): same, Accuracy. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationAccuracy↑ Higher is better
—Ihmig 2020 best

Ihmig 2020 best (SB, 10 s)

Evaluation details · caveat

Site: same

Not ranked: Window CV leaks subjects; HR1 60 s gives 0.909; 3-level 0.744.

0.898

ecg-spider-clip: balanced accuracy (Task 1, test-exclusive Rest_2)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): ecg-spider-clip, Balanced accuracy (HAR / class-imbalance). Reported rankings; evaluation protocols may differ.
RankMethod & evaluationBalanced accuracy (HAR / class-imbalance)↑ Higher is better
1TabPFN
Evaluation details · caveat

Site: chest ECG + resp, palm EDA (PLUX BITalino)

Caveat: Baseline normalization excludes the test segment.

0.85 ± 0.05

Jang: AUROC (All features)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): Jang, Area under ROC. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationArea under ROC↑ Higher is better
1XGBoost
Evaluation details · caveat

Site: wrist

Caveat: Not subject-independent; 3,969 day points, 254 with panic; accuracy 0.919.

0.905 (95% CI 0.885–0.924)
2GradientBoost
Evaluation details · caveat

Site: wrist

Caveat: Same.

0.826
3Random forest
Evaluation details · caveat

Site: wrist

Caveat: Same.

0.730

same: AUROC (Lifelog only)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): same, Area under ROC. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationArea under ROC↑ Higher is better
—XGBoost
Evaluation details · caveat

Site: wrist

Not ranked: The wearable-only baseline to beat; clinical-only 0.900, daily-log 0.827, top-10 features 0.903.

0.745 (0.693–0.796)

WESAD: accuracy (Binary stress, wrist BVP)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): WESAD, Accuracy. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationAccuracy↑ Higher is better
—Schmidt 2018 LDA
Evaluation details · caveat

Site: wrist (E4)

Not ranked: Eval-only dataset.

0.8583 (F1 0.8308)

same: accuracy (Binary stress, wrist physio)

Panic attacks and sympathetic arousal (detection, forecasting, continuous level, recovery): same, Accuracy. Reported rankings; evaluation protocols may differ.
RankMethod & evaluationAccuracy↑ Higher is better
—Schmidt 2018 RF
Evaluation details · caveat

Site: wrist (E4)

Not ranked: Same.

0.8833 (F1 0.8610)