sklearn.metrics.multilabel_confusion_matrix
Compute class-wise (default) or sample-wise (samplewise=True) multilabel confusion matrix
MCM
In multilabel confusion matrix , the count of true negatives is MCM[:, 0, 0], false negatives is MCM[:, 1, 0], true positives is MCM[:, 1, 1] and false positives is MCM[:, 0, 1].
confusion_matrix: array([[TN, FP], [FN, TP]]])
0, 0 -> TN （ただしnp.ndarray。ほかも同じ）
0, 1 -> FP
1, 0 -> FN
1, 1 -> TN
例（TODO）
Multilabel-indicator case:
Multiclass case: