Medical Text Classification Using Hybrid Deep Learning Models with Multihead Attention

Table 4

Analysis of results with different dropout rates for the proposed quad channel hybrid model.

Dropout rate	Dataset	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
0.2	Hallmarks	69.75	61.65	62.89	62.26
	AIM	90.17	87.38	85.12	86.23
0.3	Hallmarks	71.76	63.63	61.79	62.69
	AIM	56.23	51.80	52.67	52.23
0.4	Hallmarks	71.19	65.99	63.18	64.55
	AIM	79.71	73.99	72.91	73.44
0.5	Hallmarks	72.93	70.95	69.67	70.30
	AIM	94.73	87.81	88.98	88.39