Table 2.

Performance comparison.

Models	Number of cycles	Training time (s)	Test time (s)	Accuracy rate	Recall rate	F1 value
CNN	3800	3594	36	0.89	0.89	0.89
LSTM	2300	33456	139	0.77	0.77	0.76
Bi-LSTM	4000	219008	1412	0.88	0.87	0.87
Bi-LSTM-attention	4000	206272	776	0.90	0.90	0.90
RCNN	1900	185709	2051	0.92	0.91	0.91
Proposed method	5000	162950	720	0.94	0.94	0.94