Adapting transformer-based language models for heart disease detection and risk factors extraction

Journal of Big Data

Table 11 BioClinicalBERT Evaluation Metrics for each risk factor indicator based on time attribute Identification

Risk factor	Time attribute	Precision	Recall	F-Score	Support
CAD	Before_DCT	0.83	0.81	0.82	422
	During_DCT	0.00	0.00	0.00	82
	After_DCT	0.64	0.92	0.76	177
Diabetes	During_DCT	0.61	0.66	0.63	267
	Before_DCT	0.72	0.75	0.73	329
	After_DCT	0.50	0.02	0.03	111
Hypertension	Before_DCT	0.76	0.79	0.77	186
	During_DCT	0.89	0.95	0.92	452
	After_DCT	0.80	0.59	0.68	170
Obese	Before_DCT	0.00	0.00	0.00	13
	During_DCT	0.00	0.00	0.00	47
	After_DCT	0.53	0.63	0.57	62
Medication	Before_DCT	0.58	0.57	0.58	1002
	During_DCT	0.39	0.08	0.13	609
	After_DCT	0.67	0.42	0.52	808
Hyperlipidemia	Before_DCT	0.81	0.75	0.78	151
	During_DCT	0.71	0.71	0.71	141
	After_DCT	0.00	0.00	0.00	33
Weighted average		0.9338	0.9403	0.9357	42946
Macro average		0.3715	0.3614	0.3505	42946
Micro average				0.9403	42946
Accuracy				0.9403	42946