While evaluating the Linear regression model, which metric do we need to consider R square or Mean Squared Error (or which needs to be given more importance)?
Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.Sign up to join this community
Aditya already mentioned rightly so in the comments RMSE, MSE or MAE is preferred while evaluating a linear regression model. So to answer your questions I will provide with the pros and cons of using RMSE and MAE and further more put forward some points to why R-squared is not preferred.
1. MAE vs. RMSE - Which metric is better ?
Ideally this depends on the loss function of your model, the most common one being L2 - euclidean distance (you can create your own loss function too). RMSE gives relatively higher weight to large errors. It does not necessarily increase with the variance of the errors but increases with the variance of the frequency distribution of error magnitudes. While, MAE is comparatively steady.
typically, it makes more sense to give a higher weight to points away from the mean in which RMSE would be preferred. In a scenario where being away from the mean by any magnitude would be penalized equally, MAE would be preferred.
RMSE can be troublesome when calculating on different sized test samples (As is the case when training a model). RMSE has a tendency to be increasingly larger than MAE as the test sample size increases ([RMSE] ≤ [MAE * sqrt(n)].
2. Is R-squared useless ? (Kinda harsh but sorry not sorry)
To keep it short and simple.