Regression Performance Evaluation Methods

Introduction

Whenever we build any model, there is always a dilemma of how to evaluate model performance. Particularly if it regression model as there are many evaluation methods are available each with its own pros and cons. Before we dive into each of these methods, let us understand what linear regression is in simpler terms.

What is Linear Regression?

Linear regression is a way of calculating relationships between two variables. As the name suggests, it is assumed that the two variables are linearly dependent on each other. To understand it in simple language, consider the height and weight of humans. Generally height and weight are linearly dependent on each other keeping other factors like age, gender etc constant. For example, take the height and weight of any 10 male athletes with age around 30 years. You will find that tall athletes have high weights while shorter ones weigh less.

How to evaluate Regression model performance?

As we understand what linear regression is, let us do a deep dive on how to evaluate regression model performance.

MSE (Mean Square Error)

RMSE (Root Mean Square Error)

MAPE (Mean Absolute Percentage Error)

R-Square

R-Square value is always between 0 and 1 i.e., 0 means worst and 1 is best.

The problem with R2 is that it doesn’t penalise additional parameters, and with use of extra parameters it always increases or remains the same. So Adjusted R2 is the solution for this problem.

Code

Conclusion

So whenever you build a regression model, you can use any of the above methods to validate your model. I prefer to use RMSE and R2 for regression model validation.