Predicting Numbers: Regression

You have seen the seven steps that build an AI system, from deciding the dish to serving it. Once a model has learned, what does it produce? Often the answer is a single number, like a price, a wait time, or a demand figure.

Why This Matters

Predicting a number is one of the most common jobs organisations hand to AI, and one of the easiest to get wrong when the number looks confident but the pattern behind it was never there. This lesson strengthens your Understanding AI foundation. By the end, you will be able to:

• Explain what regression does and the kind of output it produces
• Identify everyday tasks where regression applies
• Describe why some numbers can be predicted and others cannot

What Regression Actually Does

Regression is AI predicting a number. It learns the relationship between inputs (the things you already know) and an outcome you want to estimate, from past examples where both were recorded. Give it the size, location, and age of many houses alongside their final sale prices, and it works out how those inputs move the price. Show it a new house, and it estimates a price. Add more useful inputs and the estimate usually sharpens.

This also tells you where regression works and where it does not. It works when a genuine, repeating relationship sits in the past data: larger homes cost more, so the pattern is learnable. It fails when no such pattern exists: no amount of data lets a model predict next week's lottery draw, because nothing in the past connects to the next one. That limit sits in the data, and no computer moves it. Events that leave no trace in history stay unpredictable, for any system, human or machine.

Key Takeaway

A confident number is not the same as a correct one.