# Part 2: Regression in Machine Learning: Predicting the Future with Data

**Previously in** [**Part 1**](https://abhis-space.hashnode.dev/part-1-what-is-supervised-learning?source=more_series_bottom_blogs), we introduced supervised learning — teaching machines with examples. Now, let’s dive deeper into **regression**, the method behind predicting things like prices, temperatures, or growth. In this post, we’ll learn what regression is, when to use it, and why it matters.

## Intro: what is Regression?

**Regression** is a type of **supervised learning** used to predict a **continuous numeric value** based on input features. The goal is to learn a mathematical relationship—or **mapping**—between the inputs (also called **independent variables**) and the output (or **dependent variable**).

### Example Problem it Solves:

> Predicting the **price of a house** based on its **size**, **location**, and **number of bedrooms**.

In this case, the model learns from past data to understand how each feature influences the house price. Once trained, it can predict the price of a **new house** given its features.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753426952090/4100aceb-8d5c-4d8f-9797-fde5959d52c6.png?auto=compress,format&format=webp align="left")

We know the **inputs** (like size and location), and the model helps us predict the **output** (price), even for examples it hasn't seen before.

## Why Regression Matters?

**Regression is essential** because it helps us make informed predictions about the future based on past data. It’s widely used in real-world scenarios where the outcome is a **numeric value**.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753429996948/eadf42b5-eca0-405d-8f5d-b81435a60072.png align="center")

* **Weather Forecasting** – Predicting tomorrow’s temperature based on historical weather data.
    
* **Sales Forecasting** – Estimating next quarter’s revenue using trends and marketing spend.
    
* **Stock Market Analysis** – Forecasting future stock prices using past price movements and indicators.
    
* **Fuel Efficiency Estimation** – Predicting how many kilometers a car can travel per litre based on engine specs and weight.
    

In short, regression gives machines the ability to **forecast, estimate, and plan** — making it a foundational tool in many industries like finance, healthcare, retail, and transportation.

## Types of Regression

### Linear Regression

* **Here, we try to draw a straight line through the data points** to best show the relationship between the input and the output.
    
* **It works better when the change in output is steady or consistent** as the input increases or decreases — this is called a **linear relationship**.
    

### Multiple Linear Regression

* **This kind of regression uses more than one input feature** to make predictions. For example, both **size** and **number of bedrooms** can be used to predict house price.
    
* Even though it still creates a **straight-line relationship**, the line exists in **multiple dimensions** — one for each input feature.
    
    ![Multiple Linear Regression](https://cdn.hashnode.com/res/hashnode/image/upload/v1753430433352/42bf5b75-60bc-4414-8713-81d1de198fe9.png align="center")
    

### Polynomial Regression

* **This one is used when the relationship between the input and output isn’t a straight line** — meaning it curves or bends.
    
* Instead of fitting a straight line, it **fits a curved line (a polynomial function)** to the data to better capture complex patterns.
    
    ![](https://cdn.hashnode.com/res/hashnode/image/upload/v1753434411522/3c846513-3484-4a9b-9fc2-b84fd3cd0106.png align="center")
    

## Key Concepts to Understand

Understanding the foundational ideas will help us to grasp any type of regression model — whether it’s linear, multiple, or polynomial.

### 1\. Features and **Targets**

**Features (Inputs / Independent Variables):** These are the variables that we provide to the model to help it make predictions.

> *Examples:* Size of the house, number of bedrooms, location score.

**Targets (Outputs / Dependent Variables):** The value that we want the model to predict, which depends on the features.

> *Example:* Price of the house.

### 2\. Residual/Error

A **residual** is the **difference between the actual value and the predicted value**. It tells us **how far off our prediction was** for a given data point.

$$\text{Residual} = y_{\text{actual}} - y_{\text{predicted}}$$

### 3\. Loss (Mean Squared Error - MSE)

![MSE - Mean Squared Error](https://cdn.hashnode.com/res/hashnode/image/upload/v1753434790615/6b18a5b3-19e7-4bb6-866b-8cc0ad566d65.png align="center")

**MSE** is a common metric used to measure how well a regression model is performing.

$$\text{MSE} = \frac{1}{n} \sum_{i=1}^{n} (y_i - \hat{y}_i)^2$$

* It calculates the **average of the squared differences** between the actual values and the predicted values (these differences are called **residuals**).
    
* Squaring the residuals ensures all errors are positive and **penalizes larger errors more heavily**.
    
* A **lower MSE** means the model's predictions are **closer to the actual values**, indicating better performance.
    

### **4\. Prediction from input features**

The model’s goal is to **learn a function** that accurately maps the **input features** to the **target output**. It improves over time by **comparing its predictions to the actual values** in the training data and **minimizing the loss function** (like Mean Squared Error). This process helps the model **adjust its internal parameters** to make better predictions as it learns from more data.

## Summary

> Regression is a supervised learning technique used to predict continuous numeric values from input features by mapping inputs to outputs. It's crucial for making informed predictions in various fields like finance, healthcare, and retail. Key types include linear, multiple linear, and polynomial regression, each suited to different relationships between input and output. Understanding features, targets, residuals, and Mean Squared Error (MSE) is essential, as these concepts help in evaluating model performance and improving predictions over time.

## What’s Next

In the [next part](https://abhis-space.hashnode.dev/part-3-mastering-linear-regression-for-accurate-predictions-the-line-function?source=more_series_bottom_blogs) of this series, we’ll take a closer look at **Linear Regression** — one of the most foundational techniques in machine learning. We’ll learn how it works, why it’s effective for predicting continuous values, and how to implement it step-by-step using Python and NumPy. We’ll break down the core math, visualize how the model fits a straight line through data, and understand how it minimizes errors to improve its predictions over time.