I have created a couple of models for my master project and I used several metrics for evaluation. I used MSE, MAE, MAPE, RMSE not because I really learned about them a lot, because I saw in many other projects these metrics being used. Now I have a problem, I need to interpret results. I search for some articles or some studies that classify metrics performance as good or bad or excellent. The only material I found now is this one.

With this interpretation

MAPE < 10% is Excellent, MAPE < 20% is Good

It would be nice to provide this formal or informal interpretation for each of the metrics MSE, MAE, MAPE, RMSE specifically for time series prediction problems.

  • 1
    $\begingroup$ Wikipedia will help you with the formal definition. As to what is a good value, it is highly dependent of your problem. $\endgroup$ May 18, 2021 at 8:33

1 Answer 1


I do not think there is a silver bullet for metric value, because it is highly problem-specific, one way or another you need to compare your result to baseline in order to understand how viable it is.

For example, in time series common baseline for the next point predicted is a previous point or some moving average. You can get an error metric like MAPE (but be careful with it, since it has its downsides) and compare your models MAPE to Baseline MAPE (say moving average) and with this comparison, you can argue about how your approach is better.

Also, check this article regarding MAPE and other metrics.


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.