According to the data on the diagonal of the confusion matrix, good samples are 7579+237+303+184=8303, total number of testing data is 9141, 8303/9141=0.908. But the Accuracy is 0.672, I want to know why.
Another question : Is the printed auc the last batch of auc? If there is no event1/2/3 in this game, auc will be 0, how to deal with it ?
Testing 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 200/200 [00:03<00:00, 53 [[7579 181 181 118] [ 119 237 9 3] [ 114 19 303 4] [ 68 7 15 184]] auc: 0.651 (auc_PR_0: 0.978 auc_PR_1: 0.602 auc_PR_2: 0.672 auc_PR_3: 0.679) Loss: 1.53e+02 Accuracy: 0.672 mAP: 0.651 Time: 3.71 s
According to the data on the diagonal of the confusion matrix, good samples are 7579+237+303+184=8303, total number of testing data is 9141, 8303/9141=0.908. But the Accuracy is 0.672, I want to know why.
Another question : Is the printed auc the last batch of auc? If there is no event1/2/3 in this game, auc will be 0, how to deal with it ?