Table 5.

Threshold-based evaluation results (dataset A).

	Balanced accuracy	Recall	Precision	F1
Decision threshold: 0.5
LR	0.672	0.597	0.177	0.274
LSTM	0.727	0.607	0.268	0.372
RF	0.618	0.288	0.340	0.312
ET	0.675	0.513	0.261	0.290
FNN	0.619	0.299	0.312	0.305
Decision threshold: optimized
LR	0.671	0.363	0.621	0.457
LSTM	0.706	0.431	0.687	0.530
RF	0.682	0.467	0.296	0.361
ET	0.689	0.493	0.282	0.358
FNN	0.618	0.269	0.448	0.336

LR: logistic regression; LSTM: long short-term memory; RF: random forest; ETs: extra trees; FNN: feedforward neural network.