What do TP, FP, FN, and TN mean?

True Positive (TP) is a correct positive prediction; False Positive (FP) is predicting positive when the truth is negative (a false alarm); False Negative (FN) is predicting negative when the truth is positive (a miss); True Negative (TN) is a correct negative prediction. In the matrix, TP and TN sit on the diagonal (correct) and FP and FN are off the diagonal (errors).

How are accuracy, precision, recall, and F1 calculated?

Accuracy = (TP + TN) / (TP + FP + FN + TN). Precision = TP / (TP + FP). Recall (sensitivity) = TP / (TP + FN). F1 = 2 × (precision × recall) / (precision + recall). Specificity = TN / (TN + FP). The tool computes all of these for you from the counts you enter, and averages them per class (macro) for multi-class matrices.

What is the difference between precision and recall?

Precision answers "of everything the model predicted positive, how much was actually positive?" while recall answers "of all the real positives, how many did the model find?" High precision minimizes false alarms; high recall minimizes misses. The F1 score combines both into a single number, which is more informative than accuracy when your classes are imbalanced.

Can I make a multi-class confusion matrix?

Yes. Click "Add class" to grow the matrix to 3×3, 4×4, or larger. The tool computes per-class precision, recall, and F1 plus the macro averages, so it works for multi-class classifiers such as image or sentiment models, not just binary ones.

Can I export the confusion matrix for a paper or report?

Yes. Download it as an SVG — a vector format that stays sharp at any size and drops straight into a manuscript, thesis, poster, slide deck, or Jupyter notebook, or opens in Illustrator or Inkscape for final edits. The metrics are drawn alongside the matrix so the figure is self-contained. Every version is free to download.

Data Science Tool

Confusion Matrix Generator with Metrics

Q: How do I make a confusion matrix?

Use the generator above: label your classes (two by default for a binary model, or add more for multi-class), then enter the counts in the grid where each row is a true class and each column is the predicted class. The matrix and its metrics — accuracy, precision, recall, and F1 — render instantly and download as a scalable SVG, free and with no sign-up.

Free confusion matrix generator — enter your classification counts for a labeled matrix with accuracy, precision, recall, and F1. Binary or multi-class.

Binary & multi-class — exact heatmap SVGAccuracy, precision, recall, F1 & specificityPer-class metrics and macro averagesDownload a publication-ready SVG — free

Enter your counts — renders an exact confusion matrix with accuracy, precision, recall & F1, free

Title

Counts — rows = true class, columns = predicted


Positive
Negative

Accuracy90.0%

Macro F189.7%

Confusion Matrix Examples

Exact engine renders — binary and multi-class, with computed metrics

View:

Binary Classifier

Exact engine render — a 2×2 binary matrix with the full metrics panel.

binaryprecision-recallf1

Multi-Class Classifier

Exact engine render — a 3×3 multi-class matrix with per-class metrics and macro averages.

multi-classmacro-averageper-class

Medical Diagnostic Test

Exact engine render — a diagnostic test where sensitivity and specificity matter.

medicalsensitivityspecificity

Sentiment Analysis

Exact engine render — a positive/neutral/negative sentiment classifier.

sentimentnlpthree-class

Image Classification

Exact engine render — a five-class image classifier heatmap.

imagedeep-learningmulti-class

Fraud Detection

Exact engine render — an imbalanced fraud model where precision and recall matter more than accuracy.

fraudimbalancedbinary

What is a confusion matrix?

A confusion matrix is a table that shows how well a classification model performs by comparing its predicted labels against the true labels. Each row is a true (actual) class and each column is a predicted class, so the cell where they meet holds the number of cases with that true label that the model predicted as that class. The diagonal holds the correct predictions; everything off the diagonal is a mistake — hence "confusion." It works for a binary classifier (a 2×2 matrix of true positives, false positives, false negatives, and true negatives) and for any multi-class problem (an N×N matrix). This generator draws the matrix and computes the standard metrics from the counts you enter.

True positives, false positives, false negatives, true negatives

True Positive (TP): the model predicted positive and the true label is positive — a correct hit.
False Positive (FP): the model predicted positive but the true label is negative — a false alarm (Type I error).
False Negative (FN): the model predicted negative but the true label is positive — a miss (Type II error).
True Negative (TN): the model predicted negative and the true label is negative — a correct rejection.

The metrics: accuracy, precision, recall, and F1

Accuracy = (TP + TN) / (TP + FP + FN + TN) — the share of all predictions that were correct. Precision = TP / (TP + FP) — of everything predicted positive, how much really was positive. Recall (also called sensitivity or the true positive rate) = TP / (TP + FN) — of all the real positives, how many the model found. Specificity = TN / (TN + FP) — the true negative rate. The F1 score = 2 × (precision × recall) / (precision + recall) is the harmonic mean of precision and recall, useful when classes are imbalanced. For a multi-class matrix, precision, recall, and F1 are computed per class and then averaged (macro average). This tool shows all of these next to the matrix so the downloaded figure is self-contained.

How to make your confusion matrix

Label your classes — keep the two default classes for a binary model, or add classes for a multi-class problem.
Enter the counts in the grid: each row is a true class and each column is the predicted class, with the correct predictions on the diagonal.
Read the metrics panel — accuracy, macro precision/recall/F1, and a per-class breakdown update as you type.
Download a clean, scalable SVG for your paper, report, notebook, or slides — free, with no sign-up.

Precision vs recall: which matters?

Precision and recall trade off against each other, and which one you optimize depends on the cost of each error. When a false positive is expensive — for example flagging a legitimate email as spam — you want high precision. When a false negative is expensive — for example missing a disease in a screening test — you want high recall (sensitivity). The F1 score balances the two, and a confusion matrix lets you see the raw counts behind them so you can judge the trade-off for your own problem rather than trusting accuracy alone, which can be misleading on imbalanced data.

Frequently Asked Questions

Try it free

View All Free Tools

Confusion Matrix Generator with Metrics

Free confusion matrix generator — enter your classification counts for a labeled matrix with accuracy, precision, recall, and F1. Binary or multi-class.

Binary & multi-class — exact heatmap SVGAccuracy, precision, recall, F1 & specificityPer-class metrics and macro averagesDownload a publication-ready SVG — free

Positive

Negative

What is a confusion matrix?

True positives, false positives, false negatives, true negatives

True Positive (TP): the model predicted positive and the true label is positive — a correct hit.

False Positive (FP): the model predicted positive but the true label is negative — a false alarm (Type I error).

False Negative (FN): the model predicted negative but the true label is positive — a miss (Type II error).

True Negative (TN): the model predicted negative and the true label is negative — a correct rejection.

The metrics: accuracy, precision, recall, and F1

How to make your confusion matrix

Label your classes — keep the two default classes for a binary model, or add classes for a multi-class problem.

Enter the counts in the grid: each row is a true class and each column is the predicted class, with the correct predictions on the diagonal.

Read the metrics panel — accuracy, macro precision/recall/F1, and a per-class breakdown update as you type.

Download a clean, scalable SVG for your paper, report, notebook, or slides — free, with no sign-up.

Precision vs recall: which matters?