Table 6.

Accuracy performance indicators for European universities’ co-publication prediction in Multidisciplinary Science.

Model	All	Exogenous	Endogenous
Best algorithm	GBM	RF	GBM
ROC	0.9418	0.9227	0.9357
Accuracy	0.8654	0.8354	0.8644
95% CI	(0.8615, 0.8692)	(0.8312, 0.8396)	(0.8605, 0.8682)
No information rate	0.8171	0.8158	0.8120
P-value [Acc > NIR]	0.0000	0.0000	0.0000
Sensitivity	0.8659	0.8308	0.8687
Specificity	0.8633	0.8560	0.8457
Pos pred value	0.9659	0.9623	0.9605
Neg pred value	0.5903	0.5332	0.5986
Prevalence	0.8171	0.8158	0.8120
Detection rate	0.7075	0.6778	0.7053
Detection prevalence	0.7325	0.7043	0.7344
Balanced accuracy	0.8646	0.8434	0.8572

Three models considered: with all, with only exogenous, and with only endogenous features.