Figure 3

Test accuracy of different models for blood donor recruitment. The three top/median performing algorithms (XGBoost, SVM, and DNN) validated in the training-test dataset were trained and evaluated by tenfold cross-validation method in different datasets. The highest accuracy scores were achieved by XGBoost in the full ABO dataset, and by SVM in the four individual blood group datasets. (A) The accuracy was calculated by XGBoost, SVM, and DNN algorithms respectively in the full ABO dataset. (B) The accuracy was calculated by XGBoost, SVM, and DNN algorithms respectively in the A blood group dataset. (C) The accuracy was calculated by XGBoost, SVM, and DNN algorithms respectively in the B blood group dataset. (D) The accuracy was calculated by XGBoost, SVM, and DNN algorithms respectively in the O blood group dataset. (E) The accuracy was calculated by XGBoost, SVM, and DNN algorithms respectively in the AB blood group dataset.