Medical Text Classification Using Hybrid Deep Learning Models with Multihead Attention

Table 6

Analysis of results with different activation functions for the proposed quad channel hybrid model.

Optimizer	Dataset (%)	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
ReLU	Hallmarks	71.92	70.92	68.62	69.75
	AIM	92.17	88.83	86.91	87.85
Sigmoid	Hallmarks	69.71	64.72	60.81	62.70
	AIM	55.31	50.90	52.21	51.54
SoftPlus	Hallmarks	70.49	64.97	66.91	65.92
	AIM	77.73	74.91	72.82	73.85
Hard sigmoid	Hallmarks	64.35	61.75	60.81	61.27
	AIM	89.48	86.22	85.22	85.71