. Author manuscript; available in PMC: 2018 Aug 1.

Published in final edited form as: J Biomed Inform. 2018 May 12;82:189–199. doi: 10.1016/j.jbi.2018.05.003

Table 4.

Imbalanced training data: test set scores with length ≤ 3 compound patterns for causative relations

Imbalance in training set	Min. Frequency: 1000			Min. Frequency: 500			Min. Frequency: 100

	Prec.	Rec.	F-score	Prec.	Rec.	F-score	Prec.	Rec.	F-score
\|N\| = 2 · \|P\|	0.744	0.724	0.732	0.868	0.775	0.819	0.865	0.846	0.855
\|N\| = 4 · \|P\|	0.851	0.698	0.766	0.922	0.760	0.833	0.924	0.837	0.878
\|N\| = 10 · \|P\|	0.950	0.646	0.769	0.967	0.745	0.842	0.967	0.816	0.885