Skip to main content

Table 1 Macro-averaging AUC, F1-score, precision, recall and accuracy of four typical classifiers based on negative samples selected by HCNS-ADR and RGNS

From: Predicting adverse drug reactions of combined medication from heterogeneous pharmacologic databases

Classifier

Negative Samples

Macro_AUC

Macro_F1

Macro_Precision

Macro_Recall

Macro_Accuracy

SVM

HCNS-ADR

0.994

0.973

0.985

0.963

0.975

SVM

RGNS

0.946

0.887

0.896

0.888

0.893

Logistic regression

HCNS-ADR

0.998

0.980

0.991

0.971

0.981

Logistic regression

RGNS

0.963

0.903

0.898

0.913

0.905

KNN

HCNS-ADR

0.983

0.920

0.972

0.883

0.936

KNN

RGNS

0.923

0.859

0.850

0.877

0.862

Random forest

HCNS-ADR

0.943

0.840

0.928

0.781

0.861

Random forest

RGNS

0.787

0.713

0.753

0.700

0.717