issues
search
KDD-OpenSource
/
DeepADoTS
Repository of the paper "A Systematic Evaluation of Deep Anomaly Detection Methods for Time Series".
MIT License
569
stars
117
forks
source link
fixing donut scores
#134
Closed
weaslbe
closed
6 years ago
weaslbe
commented
6 years ago
Fixing unstable AUC score behaviour of Donut
not thresholding the scores in predict anymore
before: AUC scores below or around 0.5
now: consistently better results for pollution and missing values
also fixing thresholding line in scores plots to use the best threshold too
auroc-1-5-2018-07-12-214753.pdf
scores_Syn Extreme Outliers (mis=0.675)-1-5-2018-07-12-214751.pdf
boxplot_auc_for_Donut_3_runs-1-5-2018-07-12-214444.pdf
auroc-1-5-2018-07-12-214446.pdf
scores_Syn Extreme Outliers (pol=0.125)-1-5-2018-07-12-214359.pdf
boxplot_auc_for_Donut_3_runs-1-5-2018-07-12-214751.pdf
weaslbe
commented
6 years ago
fixed by new master :)