Accuracy
Domain: Evaluation
🎯 Motivation
The simplest metric: what percentage of predictions are correct. Good for balanced data, misleading for imbalanced data.
📋 Concrete Example
100 questions, 90 correct. Accuracy = 90%. But if 90 questions are yes and the model always says yes, it gets 90% accuracy without being smart. Use other metrics too.
🔗 Analogy
A student’s grade percentage - 90/100 correct = 90%. Simple and intuitive. But it does not tell you where they struggled. Accuracy alone can be misleading.