Table 1.

Here, we present results to compare prediction performance for different lengths of time intervals when predicting flux increases.

Prediction horizon (min)	Accuracy	Sensitivity	Specificity	Precision	AUC
1	0.731	0.659	0.732	0.023	0.753
2	0.706	0.620	0.708	0.040	0.713
3	0.689	0.590	0.692	0.053	0.687
4	0.678	0.560	0.683	0.065	0.667
5	0.667	0.546	0.673	0.075	0.652

Values in bold indicate the best score for the same indicator. With the exception of the precision metric, performance improved as the prediction time becomes shorter. We confirmed statistically significant differences in terms of AUC using t-tests among all problem settings.