I now develop multiple linear regression model and I want to check the model accuracy. i found that it is better to use some accuracy measurement namely, Mean Absolute Error (MAE), the Mean Absolute Percentage Error (MAPE) and the Root Mean Square Error (RMSE).
However, I am not quite the difference between MAE and RMSE. Any suggestions please?
Hello, Chea Sovann
Look at this post.
Http://www.geosci-model-dev.net/7/1247/2014/gmd-7-1247-2014.пдф
Each of the errors has a place in the analysis.
If you compare results, it is preferable to compare percentages rather than absolute values.
M. Kaneva
The difference is in a term of formula used. However if we want to compare 3 models using MAE, MSE, RMSE and MAPE, the value of MAE, MSE, RMSE and MAPE will show the least value for the best model.
hi
Every tool has advantages and disadvantages,for example, RMSE it's best tool for showing bigger deviations and helps to provide a complete picture of the error distribution. MAEMAE is the most natural measure of average error magnitude and MAPE it's useful to compare the precision between different volumes in study
for more information , see attached links
Raed
http://www.int-res.com/articles/cr2005/30/c030p079.pdf
http://www.geosci-model-dev.net/7/1247/2014/gmd-7-1247-2014.pdf
Hi Chea,
Error analysis really depends on the individual, that is, what are you interested in exposing in the data. From your question, this is a matter of "squared or absolute". When you employ any of the "absolute" techniques, there is a proportional weight assigned to each error term. On the other hand, because of the square in RMSE, error terms with larger magnitudes will be more pronounced (making them even larger). This might be what you want, for example, if you are concerned about the impact of large errors as opposed to the smaller ones. In the same understanding, "absolute" based error analysis will be more robust against outliers in your data sets.
Modisa
Hi Chea.
MAE y MAPE are measures that indicates about the mean of the dispersion between predicted and observed value, for each one with the linear model (absolute difference). RMSE is a measure of model error, it is more complet (it is my opinion). Both are useful to evaluate a LRM.
The difference is the usage of difference formula. However, the conclusion for the best method will be the same if we compare 2 or more methods using the 3 error formula.
[Accuracy measures for a forecast model — accuracy • forecast (robjhyndman.com)](https://pkg.robjhyndman.com/forecast/reference/accuracy.html)
This can be helpful to know the differences between different evaluation metrics and how to calculate them https://www.reneshbedre.com/blog/multiple-linear-regression.html
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling