Skip to content
Exit presentation

Three Kinds of Guess, and Their Data

Guessing a number from past data

A cycle shop prices used cycles. Older cycles sell for less. Can a program learn that?

Six used cycles. Each dot is one cycle.
Step 1 of 4
  1. Six used cycles. Each dot is one cycle.
  2. A line is drawn. It passes near the dots.
  3. A new cycle is 4 years old.
  4. Read the line at age 4. The guess is about ₹3,400.

Yes, a program can learn a line from past prices. It reads a new price from that line.

Regression is a way to guess a number from past data. The answer can be any number in a range, like a price. A range runs from a low number to a high one.

Here the line slopes down: it goes lower as age rises. A cycle two years older gets a guess about ₹930 lower.

Regression learns from past numbers and guesses a new number.

Watch out!

Trust the line less far from the dots. The dots cover ages 1 to 7. Ages 0 and 8 lie outside them. There the line had no dots to learn from. So the guess is weaker.

Slide the age. Find the age for each target price.

Loading lab…

Jump to lesson

This track

  1. 0 Start here
  2. 1 Guessing a number from past data
  3. 2 Making groups when nobody gave labels
  4. 3 Named group, number or found groups?
  5. 4 What a dataset holds
  6. 5 Structured, unstructured or both?
  7. 6 Train, check, test
  8. 7 The big picture
  9. 8 Track quiz

Other tracks