Accuracy Calculator
Measure how often your classification model predicts correctly by analyzing true positives, true negatives, false positives, and false negatives.
Accuracy
83.33%
Correct predictions / Total observations
Error Rate
16.67%
Incorrect predictions / Total observations
Dataset Size
240
TP + TN + FP + FN
| Metric | Value | Interpretation |
|---|---|---|
| True Positives (TP) | 120 | Correctly predicted positives |
| True Negatives (TN) | 80 | Correctly predicted negatives |
| False Positives (FP) | 15 | Incorrect positives (Type I error) |
| False Negatives (FN) | 25 | Incorrect negatives (Type II error) |
| Precision | 88.89% | Share of positive predictions that were correct |
| Recall (Sensitivity) | 82.76% | Share of actual positives that were detected |
| Specificity | 84.21% | Share of actual negatives that were correctly excluded |
| F1 Score | 85.71% | Harmonic mean of precision and recall |
How to Use This Calculator
Collect confusion matrix counts
Summarize model predictions into TP, TN, FP, and FN values from your evaluation dataset.
Enter each value
Fill in all confusion matrix fields. Use whole numbers or decimals if you are averaging multiple folds.
Review accuracy and supporting metrics
Accuracy, error rate, precision, recall, specificity, and F1 score update instantly.
Compare with your objectives
Determine whether performance meets project requirements or if further model tuning is needed.
Formula
Accuracy = (TP + TN) ÷ (TP + TN + FP + FN)
Error Rate = (FP + FN) ÷ (TP + TN + FP + FN)
Precision = TP ÷ (TP + FP) • Recall = TP ÷ (TP + FN) • Specificity = TN ÷ (TN + FP)
F1 Score = 2 × (Precision × Recall) ÷ (Precision + Recall)
Accuracy summarizes the share of total predictions that were correct. Combine it with precision, recall, and specificity to understand trade-offs in imbalanced datasets.
Full Description
Accuracy is a fundamental metric for evaluating classification performance. It measures the proportion of correct predictions across all classes by summing true positives and true negatives and dividing by the total number of observations. While easy to interpret, accuracy alone can be misleading when class distributions are imbalanced—high accuracy may hide poor performance on minority classes.
To provide context, this calculator also reports precision, recall, specificity, and F1 score so you can assess errors in greater detail. For instance, a medical diagnostic model may require high recall to minimize missed cases, whereas a spam filter may prioritize precision to avoid flagging legitimate emails. Accuracy offers the broad overview, but supporting metrics illuminate model behavior in mission-critical scenarios.
Use this tool for quick validation after training models, comparing experiments, or explaining performance to stakeholders. Pair the outputs with confusion matrix visualizations to diagnose which labels need improvement and how class weighting or threshold tuning could help.
Frequently Asked Questions
When is accuracy a reliable metric?
Accuracy works well when classes are balanced and the cost of false positives and false negatives is similar. For imbalanced or high-stakes problems, inspect precision, recall, and specificity alongside accuracy.
Can I use probabilities instead of counts?
Yes. If you average results across cross-validation folds or bootstrap samples, you can enter fractional values—the formulas still apply because they operate on rates.
How does accuracy relate to balanced accuracy?
Balanced accuracy averages recall across classes to account for imbalance. Our calculator reports standard accuracy; compute balanced accuracy by averaging recall for each class separately.
What accuracy threshold should I target?
Benchmarks depend on context. Compare against baseline models (e.g., always predicting the majority class) and business requirements to determine an acceptable target.
Why does F1 score show a dash?
F1 requires both precision and recall. If either value is undefined because the denominator equals zero, the calculator displays an em dash to signal insufficient data.