Regression metrics play the same role for machine learning models that marks and grades play for students in exams.

During training, the model learns from mistakes.
During testing, the model is not taught; it is evaluated.
Just like exams give marks, regression metrics give scores to models.
These scores help us decide how good the model really is.

Regression Performance Metrics

What Are Regression Performance Metrics?

In regression, a machine learning model predicts numerical values such as:

• house price
• marks
• salary
• GPA
• delivery time

After training the model, we test it on new (unseen) data.
Now the big question is:

How accurate are the predictions?

To answer this objectively, we use regression performance metrics.

 Definition

A regression performance metric is a mathematical measure that quantifies how close the predicted values of a regression model are to the actual (real) values.

---

 Actual Value vs Predicted Value (Foundation)

• Actual value (y):
  The real value already known from data (ground truth)
• Predicted value (ŷ):
  The value predicted by the model

 Error for one data point:

Error=y−y^

Metrics combine errors from all test data points into a single meaningful number.