Confusion Matrix
Table showing correct and wrong predictions of a classification model according to classes.
Mixed matrix (confusion matrix) is a table that shows what classes of the model by comparing the predictions of a classification model with real labels, which are incorrect in lines and columns. There are four basic cells in double classification: real positive, false positive (corching alarm), real negative and false negative (complexed case). This is the basis of the calculation of almost all classification metrics such as four numbers, precision, remember, F1 score and accuracy.
The strength of the mixed matrix comes from revealing the nature of the error instead of a single summary number: for example, in cases where false negatives in a disease diagnostic model (existing pregnancy) are much more dangerous, the mess matrix clearly indicates this imbalance. In multi-class problems, the matrix NxN size expands and is presented in the form of heat map (heatmap) to visualize which classes are often mixed with each other, which gives developers concrete tips on which classes should be more data or property engineering.
The strength of the mixed matrix comes from revealing the nature of the error instead of a single summary number: for example, in cases where false negatives in a disease diagnostic model (existing pregnancy) are much more dangerous, the mess matrix clearly indicates this imbalance. In multi-class problems, the matrix NxN size expands and is presented in the form of heat map (heatmap) to visualize which classes are often mixed with each other, which gives developers concrete tips on which classes should be more data or property engineering.