Confusion Matrix Knowledge:
- Precision (precision):
- Recall (recall):
- F1-Score (P and R of the harmonic mean):
- To help understand the confusion matrix diagram (Case This figure reflects the recognition of handwritten numeral 5):
ROC / AUC curve:
- The curve is a common binary model evaluation criteria, even more than the usual confusion matrix related indicators.
- The horizontal axis represents the graph FPR (False Positive Rate), the vertical axis represents the TPR (True Positive Rate).
- Graph showing ROC (receiver operating characteristic), AUC is the area under the curve area value.
- Generally, the AUC area close to 1 as possible (closer to the upper left of the ROC curve), but close to 0.5 worse (diagonal of the ROC curve of FIG closer)
- ROC / AUC curve schematic:
See related python code that implements a special template article.