อ่านเมทริกซ์ความสับสน

The confusion matrix is a small table showing every prediction sorted by what was true and what you guessed. It reveals exactly where errors live.

Rows are the true labels and columns are the predicted ones. Each cell counts how many examples fell into that true-versus-predicted combination.

For two classes you get four counts: true positives, true negatives, and the two error types. Every metric you know is built from these.

A true positive is a real positive you correctly flagged, like real spam sent straight to the spam folder. A clean win.

A true negative is a real negative you correctly left alone, like a normal email that stays safely in the inbox.

A false positive is a false alarm: you flagged something that was actually negative, like a real email wrongly marked as spam.

A false negative is a miss: a real positive you let slip through, like spam that lands in the inbox unflagged.

Correct predictions sit on the diagonal. A strong model concentrates counts there and leaves the off-diagonal cells nearly empty.

scikit-learn assembles the matrix from your true and predicted labels in a single call.

A plotted matrix makes errors jump out, so you instantly see which class gets confused for which.

Read off-diagonal cells to find the model blind spot. A large false-negative count, for example, warns you that real positives are being missed.

Name the error type from the description.

The confusion matrix splits results into four cells. The diagonal holds wins; off-diagonal cells expose exactly where your model errs. 🔍