Table 2.

Performance of several models trained using classical machine learning methods and federated learning methods, where the number of participating clients in the federation is n = 2, tested on the PDBP dataset

Algorithm name	ROC-AUC	AUC-PR	Balanced accuracy	Precision	Recall	F 0.5	F 1	F 2	Log loss	Matthews correlation coefficient
AdaBoost classifier	0.834 ± 0.021	0.891 ± 0.015	0.697 ± 0.026	0.757 ± 0.023	0.917 ± 0.033	0.835 ± 0.016	0.834 ± 0.009	0.905 ± 0.004	0.639 ± 0.005	0.456 ± 0.029
Bagging	0.812 ± 0.01	0.871 ± 0.01	0.696 ± 0.019	0.753 ± 0.015	0.932 ± 0.015	0.828 ± 0.007	0.84 ± 0.004	0.903 ± 0.003	1.291 ± 0.226	0.463 ± 0.027
GradientBoosting classifier	0.856 ± 0.013	0.9 ± 0.016	0.716 ± 0.018	0.766 ± 0.014	0.938 ± 0.013	0.856 ± 0.007	0.857 ± 0.003∗	0.908 ± 0.003∗	0.572 ± 0.042∗	0.502 ± 0.024
KNeighbors classifier	0.586 ± 0.024	0.735 ± 0.019	0.551 ± 0.019	0.664 ± 0.011	0.946 ± 0.024	0.707 ± 0.008	0.783 ± 0.001	0.899 ± 0.0	3.322 ± 0.641	0.169 ± 0.042
LinearDiscriminantAnalysis classifier	0.702 ± 0.012	0.794 ± 0.008	0.64 ± 0.013	0.734 ± 0.01	0.776 ± 0.01	0.622 ± 0.305	0.661 ± 0.324	0.751 ± 0.368	2.104 ± 0.19	0.288 ± 0.024
LogisticRegression classifier	0.771 ± 0.011	0.842 ± 0.009	0.657 ± 0.008	0.73 ± 0.005	0.901 ± 0.01	0.791 ± 0.004	0.81 ± 0.005	0.901 ± 0.001	0.996 ± 0.039	0.368 ± 0.02
MLP classifier	0.671 ± 0.013	0.826 ± 0.007	0.619 ± 0.012	0.711 ± 0.008	0.839 ± 0.01	0.749 ± 0.009	0.789 ± 0.007	0.899 ± 0.001	8.313 ± 0.708	0.265 ± 0.026
QuadraticDiscriminantAnalysis classifier	0.525 ± 0.022	0.721 ± 0.022	0.525 ± 0.022	0.671 ± 0.024	0.366 ± 0.097	0.688 ± 0.0	0.779 ± 0.0	0.898 ± 0.0	18.716 ± 1.33	0.05 ± 0.042
RandomForest	0.736 ± 0.006	0.825 ± 0.005	0.524 ± 0.005	0.649 ± 0.003	0.985 ± 0.005∗	0.764 ± 0.007	0.792 ± 0.005	0.899 ± 0.0	0.596 ± 0.004	0.132 ± 0.025
SGD classifier	0.662 ± 0.017	0.845 ± 0.007	0.65 ± 0.016	0.728 ± 0.011	0.878 ± 0.024	0.758 ± 0.01	0.803 ± 0.007	0.898 ± 0.0	10.11 ± 0.525	0.343 ± 0.034
SVC classifier	0.701 ± 0.007	0.808 ± 0.004	0.593 ± 0.011	0.693 ± 0.006	0.844 ± 0.02	0.742 ± 0.004	0.793 ± 0.002	0.901 ± 0.001	0.65 ± 0.019	0.214 ± 0.029
XGBoost classifier	0.862 ± 0.008∗	0.905 ± 0.007∗	0.719 ± 0.021	0.77 ± 0.016	0.932 ± 0.013	0.864 ± 0.006∗	0.857 ± 0.003∗	0.906 ± 0.003	0.691 ± 0.031	0.504 ± 0.03
XGBoost Random Forest classifier	0.829 ± 0.007	0.89 ± 0.006	0.732 ± 0.02∗	0.781 ± 0.016∗	0.918 ± 0.01	0.849 ± 0.003	0.855 ± 0.002	0.905 ± 0.003	2.715 ± 0.254	0.515 ± 0.031∗
FedAvg LR	0.665 ± 0.128	0.826 ± 0.011	0.565 ± 0.052	0.673 ± 0.028	0.96 ± 0.032∗	0.745 ± 0.045	0.794 ± 0.012	0.899 ± 0.002	0.829 ± 0.108	0.187 ± 0.147
FedAvg MLP	0.69 ± 0.018	0.78 ± 0.012	0.629 ± 0.01	0.719 ± 0.007	0.828 ± 0.022	0.744 ± 0.008	0.791 ± 0.007	0.899 ± 0.001	1.038 ± 0.239	0.282 ± 0.024
FedAvg SGD	0.775 ± 0.011	0.847 ± 0.008	0.689 ± 0.011	0.77 ± 0.008∗	0.8 ± 0.01	0.794 ± 0.005	0.809 ± 0.004	0.902 ± 0.002	0.559 ± 0.013∗	0.385 ± 0.023
FedAvg XGBRF	0.794 ± 0.007∗	0.876 ± 0.009∗	0.695 ± 0.023∗	0.754 ± 0.017	0.919 ± 0.012	0.825 ± 0.007∗	0.838 ± 0.008∗	0.902 ± 0.003∗	0.691 ± 0.0	0.451 ± 0.035∗
FedProx μ = 0.5 LR	0.704 ± 0.101	0.823 ± 0.015	0.584 ± 0.042	0.683 ± 0.022	0.943 ± 0.03	0.762 ± 0.038	0.795 ± 0.009	0.9 ± 0.002	0.866 ± 0.092	0.232 ± 0.115
FedProx μ = 0.5 MLP	0.7 ± 0.008	0.791 ± 0.006	0.63 ± 0.011	0.719 ± 0.007	0.833 ± 0.016	0.748 ± 0.007	0.794 ± 0.004	0.899 ± 0.001	1.312 ± 0.124	0.284 ± 0.023
FedProx μ = 2 LR	0.761 ± 0.008	0.835 ± 0.007	0.601 ± 0.005	0.691 ± 0.003	0.947 ± 0.013	0.787 ± 0.008	0.804 ± 0.003	0.9 ± 0.001	0.875 ± 0.014	0.293 ± 0.01
FedProx μ = 2 MLP	0.695 ± 0.022	0.785 ± 0.015	0.631 ± 0.02	0.722 ± 0.013	0.818 ± 0.02	0.747 ± 0.014	0.791 ± 0.005	0.899 ± 0.001	1.231 ± 0.285	0.282 ± 0.044

Data reported are mean and standard deviation across K = 6-fold cross-validation. Best performing algorithms for each metric are indicated by an asterisk.