0

I have a dataset having values ranges from 200 to 300. After the data are processed by an algorithm (output will be correlated to input), RMSE is calculated between the input values and the output values obtained from the algorithm. The RMSE value obtained is 23.33. How can I say/find the RMSE value obtained is good or high? Is the RMSE result has a relation with the amplitude of the data?

Tina
  • 1
  • RMSE is usually used to test algorithm if it predicts good enough.Generally speaking the less the value is the better. I mean it is calculated between output and actual values. Not sure whats the goal to use it with input and output, though. – Many Feb 12 '21 at 10:37
  • 2
    See [this](https://datascience.stackexchange.com/a/9173/111290) and [this](https://stats.stackexchange.com/q/242787/159540). – David M. Feb 12 '21 at 10:46
  • 1
    The RMSE is relative to a reference estimating procedure, rather than absolute, except in the rare cases a Cramèr-Rao lower bound exists. – Xi'an Feb 12 '21 at 11:16
  • 1
    Calling something "good" or "high" is a value proposition, not a statistical question. Suppose I measured a bunch of lengths varying from 200 to 300 and got an RMSE of 23.33. Is that good or bad? Well, if I'm measuring the lengths of highway sections in meters or lengths of cattle in centimeters, that might be a good RMSE; but if I am measuring the contents of cans of food in grams, it may be a terrible RMSE. – whuber Feb 12 '21 at 15:13
  • 1
    The proposed duplicate is in terms of MAPE and sMAPE, but the exact same points apply. – Stephan Kolassa Feb 12 '21 at 15:23

0 Answers0