Skip to content
PlayOwl

Computers & Digital Literacy · Grade 9

Training, Validation and Test Sets

Learners split data into training, validation and test sets and compute the counts for common splits. They use validation results to choose settings such as k and tree depth, read the overfitting signature in training and validation error, and learn why the test set is used only once. They then hunt for data leakage in fictional project logs and choose stratified, time-based or group splits when a simple random split would mislead.

  • 5 lessons
  • About 120 minutes
  • Splits and leakage check: 10 questions
  • Ways to answer: fill in the blanks, choose an answer, trace code, highlight a text, match pairs and type a number
Try it now (no account) One of the five free starter courses Open in PlayOwl

What learners will discover

  • Explain the purposes of the training, validation and test sets and compute split counts for common splits such as 70/15/15 and 60/20/20.
  • Use validation results to choose a setting such as k or tree depth, and explain why the test set is used once, at the end.
  • Recognize underfitting and overfitting from training and validation error, including the signature of training error falling while validation error rises.
  • Identify data leakage (duplicates across splits, future information, preprocessing on all data, features derived from the label) and describe the fix.
  • Choose stratified, time-based or group splits when a situation requires them, and compute stratified counts.

Lessons

  1. Why split, and how big 7 steps4 practiceabout 22 min
  2. Choosing settings with validation data 6 steps4 practiceabout 26 min
  3. Underfitting and overfitting 5 steps3 practiceabout 22 min
  4. Data leakage 5 steps3 practiceabout 26 min
  5. Stratified, time-based and group splits 7 steps4 practiceabout 24 min
  6. Splits and leakage check 10 questions at the end

Keep exploring

More 9th grade computer skills

Guide for grown-ups