Regression vs Classification

Before choosing a machine learning model, it is essential to understand the type of problem we are trying to solve.

In supervised learning, problems are generally divided into regression and classification.

---

Regression

Regression problems aim to predict a continuous numerical value.

The target variable can take any value within a range, and the model learns a quantitative relationship between the input features and the target.

Examples of regression problems:

• House price prediction
• Temperature forecasting
• Energy consumption estimation

Output:

• A real number (e.g. 245000, 18.7, 0.92)

---

Classification

Classification problems aim to assign an input to a discrete class.

The target variable represents categories, and the model learns to distinguish between different classes based on the input features.

Examples of classification problems:

• Spam vs non-spam email detection
• Cat vs dog image classification
• Disease detection (yes / no)

Output:

• A class label (e.g. “spam”, “cat”)
• Or class probabilities

---

Key Differences

Aspect	Regression	Classification
Target type	Continuous	Discrete
Output	Numeric value	Class or probability
Typical metrics	MSE, RMSE, R²	Accuracy, Precision, Recall
Common models	Linear Regression, Random Forest	Logistic Regression, KNN

---

Why This Distinction Matters

Choosing the wrong problem type leads to: - incorrect model selection - inappropriate evaluation metrics - misleading results

Understanding whether a problem is a regression or a classification task is the first fundamental step in any machine learning project.