AI Terminology Course
AI Terminology
/
Beginner

Accuracy

Definition

The most fundamental evaluation metric in Machine Learning. It is the percentage of total predictions that the AI model got completely correct.

Explain Like I'm New

If you take a 100-question test and get 90 questions right, your Accuracy is 90%.

Real World Example

An AI is built to predict if an email is spam or not. Out of 1,000 emails, it correctly categorized 950 of them. The Accuracy of the model is 95%.

Common Use Cases

  • •General model evaluation
  • •Balanced datasets

Interview Questions

basic

  • If an AI guesses correctly 80 times out of 100, what is its accuracy?

intermediate

  • Why is Accuracy a terrible, misleading metric if your dataset is highly imbalanced?

Flash Cards

Question

What is the accuracy?

Click to reveal answer
Answer

80%.

Question

Why misleading?

Click to reveal answer
Answer

Imagine a hospital where 99% of people are healthy and 1% have a rare disease. You build a broken AI that just prints 'Healthy' every single time, without even looking at the patient. That broken AI will be mathematically 99% Accurate, even though it is completely useless at finding the disease.