Medical Text Classification Using Hybrid Deep Learning Models with Multihead Attention

Table 2

Analysis with different convolution kernel conditions for the proposed quad channel hybrid model.

Kernel filter sizes	Dataset	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
[1, 3]	Hallmarks	62.58	55.76	54.82	55.28
	AIM	81.44	75.36	79.21	77.23
[1, 3, 5]	Hallmarks	72.18	70.52	69.69	70.10
	AIM	94.12	85.71	83.93	84.81
[1, 3, 5, 7]	Hallmarks	71.49	62.98	64.98	63.96
	AIM	79.90	73.98	72.81	73.39
[1, 3, 5, 7, 9]	Hallmarks	64.71	59.37	55.91	57.58
	AIM	57.36	52.90	54.27	53.57