DataTerminology.comNEW
A table summarizing counts of correct and incorrect predictions across classes, organizing true/false positives and negatives.
DESCRIPTION
A confusion matrix cross-tabulates predicted versus actual classes, offering a granular view of classification performance. It enables computation of precision, recall, specificity, F1, and overall accuracy. For imbalanced data, per-class metrics derived from the matrix are more informative than accuracy alone. Extensions include normalized matrices and multi-label variants. In production, monitoring the matrix over time can reveal drift and degradation.
ALTERNATIVE DEFINITIONS
Wikipedia
In the field of machine learning and specifically the problem of statistical classification, a confusion matrix, also known as an error matrix, is a specific table layout that allows visualization of the performance of an algorithm.
Highlights its role as a visualization for classification performance.
Google Developers
A confusion matrix is a table that shows the number of correct and incorrect predictions made by a classifier compared with the actual outcomes.
Concise, practitioner-focused definition tied to classifier outcomes.
scikit-learn
Compute confusion matrix to evaluate the accuracy of a classification.
Defines it operationally as a computed metric in a standard ML library.
RELATED TERMS
Precision Recall F1 Score True Positive Type I Error Accuracy Calibration
COMPETING TERMS
ROC Curve Precision-Recall Curve
RELATIONS
- Broader: Matrix
KNOWLEDGE AREAS
AI Data Science Testing & Validation MLOps
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Data Analyst
EXTERNAL SOURCES
- Wikipedia - Confusion matrix
- scikit-learn - Guide to confusion matrices and metrics
- Microsoft Learn - Azure ML classification metrics overview
- Google Developers - ML Crash Course on confusion matrix
TOP QUESTIONS
Why use a Confusion Matrix instead of accuracy alone?
It reveals class-wise errors, enabling calculation of precision, recall, and specificity, which are crucial for imbalanced datasets.
Category: Relevancy
How do you interpret the cells of a Confusion Matrix?
Rows often represent actual classes and columns predicted classes (or vice versa). Diagonal entries are correct predictions; off-diagonal entries are misclassifications.
Category: Explanation
What is a common Confusion Matrix mistake?
Mixing up row/column orientation or class labels, leading to incorrect precision/recall calculations.
Category: Mistakes
How does class imbalance affect Confusion Matrix interpretation?
Accuracy can be misleading; focus on per-class recall, precision, and macro-averaged metrics.
Category: Risks
What metrics derive from a Confusion Matrix?
Precision, recall, F1, specificity, false positive rate, and balanced accuracy are computed from its entries.
Category: Explanation
How does thresholding relate to the Confusion Matrix?
Changing the decision threshold alters TP, FP, TN, FN counts, shaping ROC and PR curves.
Category: Guidance
How to visualize a Confusion Matrix?
Use heatmaps with normalized percentages per row or column to compare performance across classes.
Category: Example
Is a Confusion Matrix applicable to multilabel problems?
Yes, but it is computed per class or by averaging schemes; multilabel extensions exist.
Category: Overview
How to monitor Confusion Matrices in production?
Track class-wise metrics over time, set alerts for drift in recall or precision, and investigate shifts in misclassification patterns.
Category: Management