Accuracy
Classification Accuracy · Also known as: Overall Accuracy, Correct Classification Rate
Accuracy is the proportion of correct predictions among the total number of predictions made by a classification model. It is the most intuitive performance metric and measures how often the classifier makes correct predictions overall, regardless of class.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use accuracy when classes are roughly balanced and you want a quick, intuitive measure of overall performance. It is suitable for balanced datasets in well-defined domains such as image classification or standard benchmarks. Avoid relying solely on accuracy in imbalanced datasets, such as fraud detection or disease diagnosis, where the minority class is rare and important.
Strengths & limitations
- Simple to compute and interpret; no hyperparameters required.
- Provides a single, intuitive metric that summarizes overall model performance.
- Works well on balanced datasets and is computationally efficient.
- Misleading on imbalanced datasets; a model predicting only the majority class can achieve high accuracy despite poor performance.
- Does not distinguish between types of errors (false positives vs. false negatives).
- Cannot be used alone to assess performance on multi-class problems with varying class importance.
Frequently asked
Why is accuracy misleading for imbalanced datasets?
In imbalanced datasets, a classifier can achieve high accuracy by simply predicting the majority class for all samples. For example, in a dataset where 99% of cases are negative and 1% are positive, a model that predicts 'negative' for everything achieves 99% accuracy but is useless for identifying the important minority class.
How does accuracy differ from precision and recall?
Accuracy measures overall correctness across all classes. Precision measures the proportion of positive predictions that were correct, while recall measures the proportion of actual positive cases that were correctly identified. These metrics focus on different aspects of classification performance.
What threshold should I use for accuracy to consider a model good?
There is no universal threshold; it depends on the application domain and data characteristics. On balanced datasets, 85-95% accuracy might be reasonable for complex tasks. Always compare accuracy to baseline models and consider the cost of different error types in your specific problem.
Sources
- Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27(8), 861-874. DOI: 10.1016/j.patrec.2005.10.010 ↗
- Powers, D. M. (2011). Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation. Journal of Machine Learning Technologies, 2(1), 37-63. link ↗
How to cite this page
ScholarGate. (2026, June 3). Classification Accuracy. ScholarGate. https://scholargate.app/en/model-evaluation/accuracy
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Balanced AccuracyModel Evaluation↔ compare
- Confusion MatrixModel Evaluation↔ compare
- F1-ScoreModel Evaluation↔ compare
- PrecisionModel Evaluation↔ compare
- Recall (Sensitivity)Model Evaluation↔ compare