TABLE 3.

Numbers of correctly or incorrectly predicted pairs for positive and negative pairs by four algorithms on KISTI test data

Algorithm	ENGs considered	No. of pairs	TP	FN	FP	TN
LR	Before	483,029 (P: 189,375) (N: 293,654)	55,998	133,377	20,203	273,451
	After		131,364	58,011	39,068	254,586
NB	Before		50,782	138,593	14,267	279,387
	After		71,020	118,355	8,775	284,879
RF	Before		82,727	106,648	57,532	236,122
	After		128,105	61,270	40,479	253,175
GB	Before		75,195	114,180	24,725	268,929
	After		136,859	52,516	39,888	253,766

Abbreviations: FN, false negatives; FP, false positives; GB, gradient boosting; LR, logistic regression; N, negative; NB, naïve Bayes; P, positive; RF, random forest; TN, true negatives; TP, true positives.