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

Journal of Big Data

Table 5 BERT evaluation metrics for each risk factor indicator

Risk Factor	Indicator	Precision	Recall	F1-score	Support
CAD	Mention	0.84	0.93	0.88	260
	Event	0.71	0.76	0.74	250
	Test	0.89	0.12	0.21	68
	Symptom	0.89	0.80	0.84	94
Diabetes	Mention	0.98	1.00	0.99	693
	A1c	0.67	0.92	0.77	65
	Glucose	1.00	0.07	0.13	42
Hypertension	Mention	1.00	0.99	0.99	548
	High bp	0.98	0.99	0.98	212
Obese	Mention	0.93	1.00	0.97	127
	Obese_BMI	0.00	0.00	0.00	9
Hyperlipidemia	Mention	0.98	1.00	0.99	240
	High chol.	0.00	0.00	0.00	8
	High LDL	0.85	0.74	0.79	31
Smoker	Smoker_never	0.93	0.96	0.94	115
	Smoker_ever	0.00	0.00	0.00	3
	Smoker_current	0.00	0.00	0.00	39
	Smoker_past	0.78	0.85	0.81	123
	Smoker_unknown	0.99	0.97	0.98	203
Medication		0.89	0.68	0.77	7225
Family history	NA	0.0000	0.0000	0.0000	13
Weighted average		0.9251	0.9373	0.9284	42946
Macro average		0.3144	0.2846	0.2736	42946
Micro average				0.9373	42946
Accuracy				0.9373	42946