DataTerminology.comNEW
Classification metric measuring the proportion of correct predictions among all predictions.
DESCRIPTION
Accuracy is a baseline metric for classification models, defined as correct predictions divided by total predictions. It is intuitive and useful when classes are balanced and misclassification costs are similar. However, it can be misleading under class imbalance, where metrics like precision, recall, F1 score, or balanced accuracy provide better insight. In multi-label and multi-class settings, accuracy variants differ in strictness. It’s commonly paired with confusion matrices and threshold analysis for holistic evaluation.
ALTERNATIVE DEFINITIONS
Wikipedia
In classification, accuracy is the proportion of correct predictions among the total number of cases examined.
Provides a clear, general statistical definition for classification.
Google Developers
Accuracy is the fraction of predictions our model got right.
Simple, practitioner-oriented definition used in ML education.
RELATED TERMS
Precision Recall Confusion Matrix ROC Curve Specificity Top-k Accuracy Error Rate
COMPETING TERMS
RELATIONS
- Narrower: Balanced Accuracy Data Accuracy Top-k Accuracy
KNOWLEDGE AREAS
Data Science AI Testing & Validation MLOps
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer BI Specialist Data Analyst
EXTERNAL SOURCES
- Wikipedia - Accuracy and precision
- scikit-learn - Accuracy scoring function documentation
- Google - ML Crash Course explanation
TOP QUESTIONS
When is Accuracy appropriate for model evaluation?
Accuracy is appropriate when classes are balanced and misclassification costs are comparable. It provides a quick baseline for overall correctness.
Category: Use
Why can Accuracy be misleading for imbalanced data?
A model predicting only the majority class can score high accuracy yet be useless. Metrics like precision, recall, and F1 score better capture minority performance.
Category: Misunderstanding
How does Accuracy relate to the confusion matrix?
Accuracy equals (TP + TN) divided by total samples. It aggregates correct predictions from the confusion matrix into a single measure.
Category: Explanation
What is Balanced Accuracy and when to use it instead of Accuracy?
Balanced Accuracy averages recall across classes, useful when class distributions are skewed. It prevents dominant classes from masking poor minority performance.
Category: Alternatives
How does threshold choice affect Accuracy?
Changing classification thresholds shifts the confusion matrix and thus Accuracy. Tuning thresholds can optimize metrics aligned with business objectives.
Category: Practice
Is Accuracy meaningful for regression?
No. Accuracy is a classification metric. Regression uses measures like RMSE, MAE, or R² to assess prediction quality.
Category: Scope
What is Top-k Accuracy?
Top-k Accuracy measures whether the true class appears among a model’s top k predictions, common in multi-class tasks like image recognition.
Category: Concept
How can you improve Accuracy in a classifier?
Improve feature quality, try better models, tune hyperparameters, address data leakage, and ensure sufficient, representative training data.
Category: Improvement
What is a typical baseline Accuracy to compare against?
A naive majority-class classifier provides a baseline. Your model should significantly exceed this to demonstrate value.
Category: Benchmark