💚

Beginner-Friendly Topic

Take your time - it's perfectly normal to re-read this topic 2-3 times. Try the interactive code editor below to run code yourself. Use the Q&A section to check your understanding before moving on. You've got this! 🚀

Testing Data

Testing data is a separate set of examples that the AI model has never seen during training. It is used to measure how well the model can perform on new, real-world data.

10 min•By Priygop Team•Updated 2026

Why We Need Testing Data

Here is a problem: if you test a model on the same data it learned from, you cannot tell if it actually learned or just memorized the answers.

Imagine a student who memorizes every answer in the textbook but cannot solve new problems on an exam. That student has not truly learned.

So in AI, we always split data into two parts:
- Training data: what the model learns from (like studying a textbook)
- Testing data: what we use to evaluate the model (like an exam with new questions)

Diagram

Loading diagram…

Deep Learning ⊂ Machine Learning ⊂ Artificial Intelligence

The Train-Test Split

A typical split is 80% for training and 20% for testing
The model never sees the testing data during training
After training, the model is tested on the held-out testing data
The test score tells you how well the model will perform on real, new data
If the model scores well on training data but poorly on testing data, it has memorized instead of learned

Train-Test Split in Code

# Splitting data into training and testing sets
import random

# Full dataset of student exam scores
all_data = [
    ([5, 70, 90], 75),
    ([2, 50, 60], 55),
    ([8, 80, 95], 88),
    ([1, 40, 50], 45),
    ([6, 75, 85], 80),
    ([3, 60, 70], 65),
    ([7, 72, 88], 78),
    ([4, 55, 65], 62),
    ([9, 85, 98], 92),
    ([0, 30, 40], 35),
]

# Shuffle and split 80% train, 20% test
random.shuffle(all_data)
split_index = int(len(all_data) * 0.8)   # 80% for training

training_data = all_data[:split_index]     # first 80%
testing_data  = all_data[split_index:]     # last 20%

print(f"Total examples: {len(all_data)}")
print(f"Training examples: {len(training_data)} (80%)")
print(f"Testing examples:  {len(testing_data)} (20%)")
print()
print("Training data is used to LEARN")
print("Testing data is used to EVALUATE")
print()
print("The model is never allowed to see the testing data during training")

Try It Yourself

Try It YourselfHTML

HTML Editor

✓ ValidTab = 2 spaces

import random

all_data = list(range(1, 21))  # 20 items: [1, 2, 3, ..., 20]
random.shuffle(all_data)

# Try different split ratios
split_ratio = 0.8  # Change this to 0.7 or 0.9 and see what happens

split_index = int(len(all_data) * split_ratio)
training = all_data[:split_index]
testing = all_data[split_index:]

print(f"Split ratio: {int(split_ratio*100)}% train / {int((1-split_ratio)*100)}% test")
print(f"Training set ({len(training)} items): {training}")
print(f"Testing set ({len(testing)} items): {testing}")

Result

Click ▶ Run to see the result

Edit the code on the left, then click Run (Ctrl+Enter)

HTML|15 lines|518 chars|✓ Valid syntax

UTF-8

Common Mistake

Warning

Never let your model see any part of the testing data during training. This is called data leakage and it will give you an overly optimistic score that does not reflect real-world performance. Always split your data before doing any processing.

Key Takeaways

Testing data is a separate set of examples that the AI model has never seen during training.
A typical split is 80% for training and 20% for testing
The model never sees the testing data during training
After training, the model is tested on the held-out testing data

Topics in This Module