A clinical text classification paradigm using weak supervision and deep representation

Table 4 Comparison results of the proposed clinical text classification paradigm

Mayo Clinic Smoking Status Classification
	Precision	Recall	F1 Score
Rule-based NLP	0.91	0.91	0.91
SVM	0.80	0.79	0.80
RF	0.82	0.81	0.81
MLPNN	0.85	0.85	0.85
CNN	0.93 ^*	0.92 ^*	0.92 ^*
Proximal Femur (Hip) Fracture Classification
	Precision	Recall	F1 Score
Rule-based NLP	0.93	0.92	0.93
SVM	0.95	0.95	0.95
RF	0.93	0.93	0.93
MLPNN	0.95	0.95	0.95
CNN	0.97 ^*	0.97 ^*	0.97 ^*
i2b2 2006 Smoking Status Classification
	Precision	Recall	F1 Score
Rule-based NLP	0.91	0.89	0.88
SVM	0.86	0.84	0.84
RF	0.85	0.84	0.83
MLPNN	0.83	0.82	0.82
CNN	0.76^*	0.81^*	0.77^*

The asterisk indicates that difference between CNN and other methods is statistically significant

ISSN: 1472-6947