Making a Prediction

Every time you open a streaming app and it recommends a show, a model just made a prediction about you. It looked at your history, processed it through thousands of learned rules, and output a guess: you will probably like this. That guess is a prediction — and learning how models make them is the first step to understanding machine learning from the inside.

From Input to Output

A machine learning model is, at its core, a mathematical function. You hand it an input — a set of numbers representing features of something — and it hands back an output number (or a category). Example: imagine a model that predicts whether an email is spam. Its input might be features like: number of exclamation marks, whether the word 'FREE' appears, how many links it contains. The model combines those features using weights — numbers it learned during training — and produces an output. If that output crosses a threshold, it predicts 'spam.' If not, it predicts 'not spam.' The key insight: the model does not look at the email the way you do. It sees only numbers, runs a calculation, and returns a number. Everything else — words, images, sound — must first be converted to numbers before a model can touch it.

Consider a concrete worked example. A model predicts tomorrow's high temperature. Its features are: today's temperature, humidity, wind speed, and month of year. It multiplies each feature by a learned weight and adds the results together (this is called a linear model). If today is 72°F with moderate humidity and low wind in July, the model might output 78°F as tomorrow's prediction. A human meteorologist uses vastly more information, but the principle — combine inputs, produce output — is the same.

Classification vs. Regression

Predictions come in two major flavors, and knowing the difference matters for every lesson in this module. Classification: the model predicts a category. Is this email spam or not spam? Is this photo a cat, dog, or bird? Will this student pass or fail? The output is a class label. Regression: the model predicts a number on a continuous scale. What will tomorrow's temperature be? How much will this house sell for? How many minutes until a bus arrives? The output is a quantity. Same idea — input features, learned weights, output — but the end result is shaped differently. Many lessons ahead will use both types, so keep them straight.