Table 1.

Classification Performance of Decision Trees on Test Sets

Test	Raw Text	NLP
Sensitivity (95% CI)	0.925 (0.883–0.954)	0.933 (0.897–0.959)
Specificity (95% CI)	0.933 (0.928–0.937)	0.969 (0.964–0.973)
PPV (95% CI)	0.650 (0.621–0.671)	0.816 (0.785–0.839)
NPV (95% CI)	0.989 (0.983–0.993)	0.990 (0.985–0.994)
Weighted precision	0.949	0.968
Weighted recall	0.932	0.964
Weighted F-score	0.940	0.966

Weighted performance scores are commonly reported in the information retrieval literature. Weighted precision is the weighted average of PPV and NPV, weighted recall is the weighted average of sensitivity and specificity, and the weighted F-score is the weighted average of precision and recall for positive cases and negative cases.

NLP = natural language processing; NPV = negative predictive value; PPV = positive predictive value.