1
$\begingroup$

My Linear Regression Model Mean Absolute Error(MAE) is 0.29 and R2 0.20 , Is this a acceptable Model ?

How can increase the r2 score ?

enter image description here

$\endgroup$
3
  • $\begingroup$ Only you can decide that, is a MAE of 0.29 acceptable to you? $\endgroup$ Commented May 26, 2022 at 6:56
  • $\begingroup$ is this model is good or bad ? $\endgroup$ Commented May 26, 2022 at 7:02
  • $\begingroup$ Impossible to fully answer this question without context regarding the problem you are trying to solve. Even if you use R square and MAE by itself, these numbers can vary by subject matter. $\endgroup$ Commented May 26, 2022 at 14:29

1 Answer 1

2
$\begingroup$

Under the usual conditions (that you probably satisfy), $R^2$ can be interpreted as the proportion of variance in $y$ explained by your model. This means that your model explains $20\%$ of the variance in $y$.

Mean absolute error is the mean amount by which your model predictions miss the observed values. This means that, on average, your model misses by $0.29$.

It’s up to you to put those numbers in context. These metrics aren’t equivalent to grades in school where being over $R^2=0.90$ is an $A$-grade and scoring $R^2=0.2$ is an $F$-grade. There are situations where the low $R^2$ (or high MAE) could be fantastic for your task, and there are situations where the high $R^2$ (or low MAE) could indicate rather pedestrian performance.

$\endgroup$
1
  • 1
    $\begingroup$ I have addressed your first question about deciding if your performance metrics indicate an acceptable model. Your second question about increasing the $R^2$ score is equivalent to asking how to do machine learning, statistical modeling, and predictive analytics, so that question is much too broad for this Stack or any other. $\endgroup$
    – Dave
    Commented May 26, 2022 at 10:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.