Build confusion matrices and compute precision, recall, F1-score, accuracy, and ROC curves with interactive threshold adjustment.
Convert actual outcomes and model predictions into a confusion matrix, then inspect precision, recall, F1-score, and accuracy. When prediction scores are available, adjust the classification threshold to see how false positives and false negatives change and how those operating points form an ROC curve. The key trap is orientation: a matrix is easy to misread unless you confirm which class is positive and whether actual labels occupy rows or columns.
A concrete input and expected output from the current implementation.
Input
True positives: 8 False positives: 2 False negatives: 2 True negatives: 8
Expected output
Confusion matrix, with actual rows and predicted columns: [[8, 2], [2, 8]]. Precision: 0.80. Recall: 0.80. F1-score: 0.80. Accuracy: 0.80.
Precision and recall are both 8 / 10. Sixteen of the 20 predictions are correct, so accuracy is 16 / 20, and equal precision and recall produce an F1-score of 0.80.
For a binary classifier, the four cells are true positives, false positives, false negatives, and true negatives. Precision is TP / (TP + FP), recall is TP / (TP + FN), F1 is the harmonic mean of precision and recall, and accuracy is (TP + TN) divided by all observations. A zero denominator makes the corresponding metric undefined.
With labeled prediction scores, the selected threshold separates predicted positives from predicted negatives. Moving it changes the four confusion-matrix counts and therefore the displayed metrics. The ROC curve records true-positive rate against false-positive rate across thresholds. Its points describe different operating choices, not repeated measurements of one fixed confusion matrix.
Not necessarily. Accuracy weights every correct prediction equally, so a dominant class can hide poor detection of a rarer positive class. Inspect the underlying counts together with precision and recall. Whether the result is acceptable depends on the consequences of false positives and false negatives in the intended use.
The threshold changes which scored observations are labeled positive. Lowering it commonly admits more predicted positives, which can increase both true positives and false positives. Raising it commonly does the reverse. There is no universally best threshold; choose an operating point based on the error tradeoff relevant to the decision.
The example and behavioral notes were checked against the browser implementation. Standards and primary references below define the relevant format, formula, or platform behavior.
Calculate percentages, percentage change, and what percent one number is of another with step-by-step breakdowns.
Calculate tips and split bills between multiple people with customizable tip percentages and rounding options.
Calculate monthly payments, total interest, and generate amortization schedules for loans and mortgages.
Calculate Body Mass Index from height and weight with category classification and healthy range guidance.
Calculate sale prices, savings amounts, and final costs after single or stacked percentage discounts.
Compute mean, median, mode, standard deviation, variance, and other descriptive statistics from a data set.