shapesinfinity.
Learn AI · Beginner · 3 min

What Is Model Accuracy?

Can you guess right every time? Neither can AI—but we measure how often it does.

The Accuracy Scoreboard
Test dataModel predictsCompare to real answerCount correct answersCalculate percentageAccuracy score

Data enters, model predicts, we count correct answers, then calculate accuracy.

goes inthe AI workscomes out

In simple words

Accuracy is like a scoreboard. It shows how many answers your AI got right out of all tries.

The real definition

Accuracy measures how often a model predicts correctly. It is the percentage of correct answers divided by total answers.

Like… Test at School

Accuracy is like your test score. If you answer twenty questions and get sixteen right, your accuracy is eighty percent.

But: Real tests measure what you learned. Accuracy only measures if predictions match the right answer.

You see it every day

on your phone

Photo Recognition App

Your phone's camera identifies objects in pictures. If it guesses correctly ninety-five times out of one hundred photos, accuracy is ninety-five percent.

out in the world

Disease Detection

A hospital uses an AI model to find illness in scans. The model was correct for eighty-eight patients out of one hundred. Accuracy is eighty-eight percent.

at work

Email Filter

Your company's AI catches spam emails. It correctly identified nine hundred seventy emails out of one thousand total. Accuracy is ninety-seven percent.

Step by step

  1. 1

    Gather test data

    Collect examples where you know the correct answer.

  2. 2

    Run predictions

    Feed the data into your model and let it guess.

  3. 3

    Compare answers

    Check each prediction against the real answer you already knew.

  4. 4

    Count and calculate

    Divide correct predictions by total predictions. Multiply by one hundred for the percentage.

Remember

Accuracy shows what percentage of predictions are correctHigher accuracy means the model works betterYou need known answers to measure accuracyAccuracy alone does not tell the whole storyPerfect one hundred percent accuracy is rarely real

Memory trick

Think 'Right ÷ Total × 100.' Count your hits, divide by all tries, multiply by one hundred.

Words

Accuracy ak-YER-uh-see
The percentage of correct predictions a model makes.
Model MAH-dul
The trained program that makes predictions from data.
Prediction pruh-DIK-shun
What the AI guesses or outputs based on input.
Test data test DAY-tuh
Examples used to measure how well a model works.
Percentage pur-SEN-tij
A number out of one hundred, shown with the % symbol.
Precision pruh-SIZH-un
The percentage of predictions that are correct when the model says yes.
Recall rih-KAWL
The percentage of actual correct cases the model finds.

Check yourself

quick check

An AI model predicts correctly seventy times out of one hundred tries. What is its accuracy?

Hint

Divide correct predictions by total, then multiply by one hundred.

quick check

A photo app identifies dog breeds correctly ninety percent of the time. What does this accuracy mean?

Hint

Accuracy is the percentage of right answers, not about perfection.

think

Why might a model with ninety-eight percent accuracy still have problems in real life?

Show a good answer

Because two percent of predictions are still wrong. This matters a lot if you make millions of predictions. Also, accuracy alone does not show which types of errors happen most.

Tell a friend

“Accuracy is how often your AI guess is right. Divide correct predictions by all predictions.”

People also ask

Is ninety percent accuracy good?

It depends on the task. For photo tags, ninety percent is great. For medical scans, it might not be safe enough.

Can accuracy be one hundred percent?

Rarely. Almost all real models make some mistakes. One hundred percent usually means your test data is too easy.

Why can't I just use accuracy to judge my model?

Accuracy ignores which errors matter most. Check precision and recall for the full picture.