There are five techniques of regression metrics to check the accuracy of regression algorithm(Loss function).
1.MAE
In mean absolute error we calculate the mean error.
The graph of mean absolute error is not differentiable.
2. MSE
Mean squared error fulfill the drawback of mean absolute error as its graph is differentiable.
In mean squared error rather than using the modulus function we calculate the square of all the errors.
Mean squared error is not robust to outliers.
3. RMSE
RMSE gives the root of the mean squared error.
It's properties are same as the mean squared error.
4. R2 Score
R2 Score calculates the difference between the mean line and the regression line.
The biggest drawback of R2 score is when irrelevant columns are added the R2 score increases.
5. Adjusted R2 Score
Adjusted R2 Score fulfill the drawback of R2 Score. As it does not increase when is irrelevant columns are added.
Comments
Post a Comment