Table 3 Performance of different machine algorithms for COVID-19 mRNA vaccine responses in SOT recipients

Model	Accuracy	Accuracy (SD)	Balanced Accuracy	Balanced Accuracy (SD)	AUC	AUC (SD)
Dose 2 (detectable vs undetectable)
LR	0.824	0.101	0.825	0.083	0.900	0.095
SVM	0.917	0.093	0.921	0.084	0.950	0.081
RF	0.606	0.203	0.583	0.200	0.694	0.219
GB	0.700	0.113	0.683	0.115	0.756	0.143
Dose 3 (positive vs detectable)
LR	0.775	0.207	0.775	0.207	0.825	0.251
SVM	1.000	0.000	1.000	0.000	1.000	0.000
RF	0.766	0.081	0.775	0.075	0.825	1.952
GB	0.733	0.161	0.750	0.158	0.925	0.160
Dose 3 (positive vs non-positive)
LR	0.714	0.162	0.730	0.173	0.845	0.129
SVM	0.795	0.133	0.802	0.161	0.933	0.118
RF	0.773	0.127	0.656	0.173	0.837	0.156
GB	0.771	0.058	0.647	0.129	0.842	0.208

Multiple machine learning models were used and compared by recursive feature elimination, with 10-fold cross-validation to assess the explanatory power of the model, with the area under the ROC curve (AUC). For all classifications, the support vector machine approach gave the highest accuracy.
LR logistic regression, SVM support vector machine, RF random forest, GB gradient boosting.

Quick links

Search