Prediction of ineffectiveness of biological drugs using machine learning and explainable AI methods: data from the Austrian Biological Registry BioReg

Table 3 Best models according to highest mean AUROC score per medication

Medication	Ineffective		Best model	Sampling strategy	Mean AUROC (95% CI)
Medication	No	Yes	Best model	Sampling strategy	Mean AUROC (95% CI)
Abatacept	212	20	Ridge classifier	RUS	0.66 (0.54–0.78)
Adalimumab	493	36	XG Boost	OVS	0.70 (0.68–0.74)
Certolizumab	150	11	SVC	OVS	0.84 (0.79–0.89)
Etanercept	530	23	RF Classifier	OVS	0.68 (0.55–0.87)
Tocilizumab	339	29	XG Boost	None	0.72 (0.69–0.77)

XGBoost, extreme gradient boosting; SVC, support vector classifier; RF Classifier, Random Forest Classifier; RUS, random undersampling; OVS, oversampling

ISSN: 1478-6362