Part 1: What is Supervised Learning?

"Teaching machines by example — so they can make smart predictions on their own."

Supervised Learning, is one of the most critical foundations of machine learning. Here we will learn what it is, how it works, and the two major types of supervised learning — Regression and Classification — explained in simple terms.

Formal Definition

Supervised learning is a type of machine learning where the model is trained on a labeled dataset, which means that each input is associated with a corresponding correct output. The goal of supervised learning is to find a function that can accurately map inputs to outputs even when new, unseen inputs are given. The function learns from the patterns it acquires from the training data.

Layman’s Definition

Let’s think of it as teaching by example. The trainer gives the machine a lot of inputs and their correct outputs. By doing so for long enough, the machine finds out some patterns, so that it can predict the correct/close output values for the new, unseen inputs based on the patterns that it has learned.

Types of supervised learnings

• Regression: a model learns to predict a continuous value, like the price of a house, based on its features. It learns this from the training data and then predict the output values for new input values.

  • Problem: You want to predict the price of a house.

  • Input Features (X):

    • Number of bedrooms

    • Size in square feet

    • Age of the house

    • Location score (e.g., proximity to schools, parks)

  • Output (Y): House price in dollars.

    We will feed the machine learning model with data like the one in the below table.

    

    Then the model learns patterns from this data. When, a new house with features like 3 bedrooms, 1200 sq ft, 8 years old, and 8.0 location score, is given it predicts the price — say, $270,000.

• Classification: a model learns to sort data into categories, like labeling emails as "spam" or "not spam". It learns this from the training data with labelled examples.

  • Problem: We want to classify whether an email is spam or not spam.

  • Input Features (X):

    • Presence of certain keywords (e.g., "win", "free", "offer")

    • Number of links in the email

    • Sender email domain

    • Use of all caps or exclamation marks

  • Output (Y): Category label → "Spam" or "Not Spam"

    The training data might look like this.

    

    The model learns from these labeled examples. Then, when a new email arrives with similar features, the model predicts whether it's spam or not.

Summary

Supervised learning is a key component of machine learning where models are trained on labeled datasets to predict outcomes based on learned patterns. This method includes two main types: Regression, which predicts continuous values like house prices, and Classification, which sorts data into categories, such as identifying emails as "spam" or "not spam." By providing numerous input-output examples, models become capable of making accurate predictions on new, unseen data.

Supervised learning powers many real-world applications like facial recognition, speech-to-text systems, fraud detection, recommendation engines etc...

What’s Next

In the next part, we’ll dive deeper into one of the supervised learning algorithms - Regression. We'll explore what regression really means in the context of machine learning, how it works behind the scenes, and why it's so widely used. We'll learn about its different types — such as Linear Regression, Multiple Linear Regression, and Polynomial Regression — and how each is suited to different kinds of problems. Likewise, we'll also look at real-world examples, visualize how the model fits data, and understand key evaluation metrics like Mean Squared Error (MSE).