Fig. 2

Performance analysis of our method, comparison of feature embedding strategies, and performance comparison of GMNN network models. A, B The AUROC and AUPR values of Adam’s LR scheme and the other four strategies under the four datasets, respectively. C, D The AUROC and AUPR values in the optimization process of a value under the grid optimization method, respectively. E, F The AUROC and AUPR values of CircDA during the epoch iteration, respectively. G The graph of the CircDA loss value in the epoch iterative training phase. H The comparison of the matrix factorization feature embedding strategy in CircDA with the other two strategies. I The comparison between GMNN network model and the other four models