Asymmetric Loss Evaluation Proportion

Use this option to specify a proportion for asymmetric loss calculations for continuous dependent variables, when evaluating the model that you are fitting.

This parameter has no impact on the model fit, but changes how evaluation criteria like RMSE and MAE are calculated. The number must be between 0 and 1, and a value of 0.5 represents symmetric loss with no effect. The specified value becomes the weight placed on positive residuals, which correspond to under-predictions. Negative residuals (over-predictions) are then weighted as one minus the specified value. For example, if over-predictions are to be weighted three times more than under-predictions, then specify 0.25.

The value is analogous to the quantile level in quantile regression.

To Specify the Proportion:

8

Type the desired value values in the text box.