What is the good RMSE (root-mean-square-error) value range to justify the efficiency of multivariate linear regression model? [closed]

Question

I'm a new bee in ML domain. Following some simple ML techniques, I build a multivariate linear regression-based ML model for predicting the computation resource usage for a system. So, when I perform the test, considering my ML model, I got the RMSE value less than 7 (approx. ~ 6.632). So, I'm a bit curious to know that, is this value is acceptable for the scientific community? Or the value is too high? Thanking you in advance.

Answer 1

Let me give you two examples having the same RMSE value:

• I'm trying to predict renting price for an apartment with renting price typically lying in range 500$-1000$. An RMSE value of 15$ could be argued to be a very low RMS error and the line of best fit would satisfy my needs.
• I'm trying to predict the price of my next family dinner, with prices ranging typically between 10$-25$. The same RMSE value of 15$ could be deemed poor and the line of best fit would probably not be satisfiable.

The interpretation of many scores in regression problems (i.e. RMSE, MAE, MSE) all depends on the domain of your problem, and what you would find acceptable. You can't really compare RMSE scores unless its derived from the same regression problem.