Table 2.

Cross-validated performance of the models (5-fold stratified CV).

Model	Accuracy	Precision	Recall	F1- score	Macro PR-AUC
Random Forest	92.6%	92%	92.6%	91.9%	0.79
CNN	93.1%	91.8%	93.1%	91.8%	0.80
CNN-BiLSTM	95.1%	94.9%	95.1%	94.7%	0.88