RMSE / MAPE interpretation of graph I am a ML beginner. For a school assignment I was told to calculate both RMSE and MAPE of some machine learning algorithm.
I have done that, but to be honest they both look very similar:


My question is, what does the fact that they look almost identical tell us?
That we didn't predict many outliers?
What would you conclude from this?
Thanks!
 A: Since the RMSE (root mean squared error) is squaring the calculated errors - this means that errors on outliers are penalised to a greater extent.
This is not the case with MAPE (mean average percentage error), and for this reason the MdAPE (median average percentage error) is commonly used in its place.
Going just on the results you have provided in your graph, the prediction performance does not look too different, but I would not necessarily use this to infer whether outliers were present or not. Instead, I would be inclined to generate a QQ plot of the data and examine whether outliers are present. This would give a better indication as to whether the model is forecasting outliers effectively or not.
If a QQ plot shows significant outliers to be present, then compare the obtained RMSE score to the mean of the test set, i.e. if the RMSE is high relative to the mean, then this indicates the prediction performance was poor. On the other hand, a low RMSE relative to the mean would indicate strong performance.
