Skip to main content

Advertisement

Table 2 Results obtained for RF and NBM2 classifiers using different class balancing strategies

From: Predictability of drug-induced liver injury by machine learning

balancing strategyclassifierMCC cvMCC val
adasynRF0.63 (0.60, 0.66)0.12
oversampled_allRF0.69 (0.65, 0.71)-0.13
oversampled_minorityRF0.69 (0.65, 0.71)-0.13
smoteRF0.63 (0.60, 0.66)0.02
smote_svmRF0.61 (0.59, 0.65)-0.09
smote_borderline1RF0.61 (0.58, 0.64)-0.04
smote_borderline2RF0.59 (0.55, 0.63)-0.07
adasynNBM20.07 (0.03, 0.10)0.02
oversampled_allNBM20.24 (0.19, 0.29)-0.02
oversampled_minorityNBM20.23 (0.19, 0.28)0.07
smoteNBM20.20 (0.15, 0.25)-0.2
smote_svmNBM20.24 (0.20, 0.29)0.1
smote_borderline1NBM20.23 (0.19, 0.29)-0.11
smote_borderline2NBM20.11 (0.06, 0.16)-0.01
  1. Boldface indicates the best performance of RF or NBM2 models either in cross validation or in validation