Artificial IntelligenceEvaluation & Metrics

Confusion Matrix

Overview

Direct Answer

A confusion matrix is a square table that summarises the performance of a classification model by displaying the frequency of true positives, true negatives, false positives, and false negatives against actual class labels. It provides the raw data foundation for calculating derived performance metrics such as precision, recall, and F1-score.

How It Works

The matrix organises predictions into four cells: correct predictions (true positives and true negatives along the diagonal) and incorrect predictions (false positives and false negatives off-diagonal). Each row represents an actual class while each column represents a predicted class, allowing direct visual inspection of where the model makes systematic errors or excels by class type.

Why It Matters

Classification metrics derived from the matrix—such as sensitivity and specificity—enable practitioners to assess whether a model's errors are acceptable for the specific business context. In domains like medical diagnosis or fraud detection, understanding the distribution of error types is critical for regulatory compliance and risk management, as false positives and false negatives carry different operational costs.

Common Applications

The matrix is fundamental in evaluating binary and multiclass classifiers across medical imaging, credit risk assessment, spam detection, and disease screening programmes. It enables comparison of model performance before deployment and supports threshold optimisation when decision boundaries must be adjusted for production constraints.

Key Considerations

The matrix assumes balanced class representation; with severe class imbalance, derived metrics can be misleading without careful interpretation. Practitioners must select appropriate evaluation metrics from the matrix's four values based on domain-specific consequences of different error types rather than relying solely on overall accuracy.

More in Artificial Intelligence

AI Safety

Safety & Governance

The interdisciplinary field dedicated to making AI systems safe, robust, and beneficial while minimizing risks of unintended consequences.

AI Ethics

Foundations & Theory

The branch of ethics examining moral issues surrounding the development, deployment, and impact of artificial intelligence on society.

Bayesian Reasoning

Reasoning & Planning

A statistical approach to AI that uses Bayes' theorem to update probability estimates as new evidence becomes available.

AI Feature Store

Training & Inference

A centralised platform for storing, managing, and serving machine learning features consistently across training and inference.

Reinforcement Learning from Human Feedback

Training & Inference

A training paradigm where AI models are refined using human preference signals, aligning model outputs with human values and quality expectations through reward modelling.

Speculative Decoding

Models & Architecture

An inference acceleration technique where a small draft model generates candidate token sequences that are verified in parallel by the larger target model.

AI Alignment

Safety & Governance

The research field focused on ensuring AI systems act in accordance with human values, intentions, and ethical principles.

AI Democratisation

Infrastructure & Operations

The movement to make AI tools, knowledge, and resources accessible to non-experts and organisations of all sizes.