Score Function in Machine Learning Explained with Examples

📘 Understanding the Cost Function (Mean Squared Error)

In machine learning, a cost function tells us how well a model is performing. One of the most common cost functions used in regression problems is Mean Squared Error (MSE).

🔍 Step-by-Step Explanation

A model makes predictions (for example, predicting a student's test score), and we compare these predictions with the actual results.

For each prediction, calculate the difference between the predicted value and the actual value. This difference is known as the error.

Squaring the error ensures all values are positive and penalizes larger mistakes more heavily.

The squared errors are averaged to produce a single value — the cost.

📐 Cost Function Formula

Cost = (1 / n) × Σ (y_i − ŷ_i)²

n: number of data points
y_i: actual value
ŷ_i: predicted value

💻 CLI Output Example

$ python calculate_mse.py Predicted: [78, 85, 90] Actual: [80, 82, 88] Errors: [-2, 3, 2] Squared: [4, 9, 4] Mean Squared Error = 5.67

💡 Key Takeaways

• Lower cost = better model performance
• Squaring errors emphasizes large mistakes
• MSE is smooth and easy to optimize
• Widely used in linear regression problems

What Is a Cost Function? Understanding Its Role in Model Training

Evaluating a Model with a Cost Function

Understanding Mean Squared Error using a house price prediction example

When building machine learning models, predictions are rarely perfect. To understand how well a model performs, we use a cost function. This page walks through a concrete example using house prices and Mean Squared Error (MSE).

Scenario Overview

Imagine you have a model that predicts house prices, and you want to evaluate how accurate those predictions are.

🏠 Actual vs Predicted Prices

House	Actual Price ($)	Predicted Price ($)
1	200,000	210,000
2	250,000	240,000
3	300,000	290,000

What Is a Cost Function?

A cost function measures how far the model’s predictions are from the actual values. In regression problems, a commonly used cost function is Mean Squared Error (MSE).

Step-by-Step Cost Calculation

1️⃣ Calculate Errors

The error is the difference between the actual price and the predicted price.

House 1: 200,000 - 210,000 = -10,000
House 2: 250,000 - 240,000 =  10,000
House 3: 300,000 - 290,000 =  10,000

2️⃣ Square the Errors

Squaring the errors ensures that negative and positive errors do not cancel each other out. It also penalizes larger mistakes more heavily.

(-10,000)² = 100,000,000
( 10,000)² = 100,000,000
( 10,000)² = 100,000,000

3️⃣ Average the Squared Errors (MSE)

To get the Mean Squared Error, we take the average of all squared errors.

MSE = (100,000,000 + 100,000,000 + 100,000,000) / 3
MSE = 100,000,000

Interpreting the Result

An MSE of 100,000,000 means that, on average, the model’s predictions deviate significantly from the actual prices.

The purpose of training a machine learning model is to adjust its parameters so this cost function is minimized. As the MSE decreases, predictions become closer to real-world values.

💡 Key Takeaways

• Cost functions quantify prediction error
• Mean Squared Error penalizes large mistakes
• Squaring prevents error cancellation
• Lower MSE indicates a better-performing model
• Training aims to minimize the cost function

Educational walkthrough of Mean Squared Error in regression models.