Regression: objectives and metrics

Name Used for optimization User-defined parameters Formula and/or description
MAE +

use_weights

Default: true

Calculation principles
MAPE +

use_weights

Default: true

Calculation principles
Poisson +
Calculation principles
Quantile +

alpha

Default:  0.5

Calculation principles

RMSE +
Calculation principles
LogLinQuantile +

alpha

Default:  0.5

Calculation principles

Depends on the condition for the ratio of the label value and the resulting value:

Lq +

q

Default: Obligatory parameter

Calculation principles
Huber +
Calculation principles

Expectile +
Calculation principles

FairLoss
Calculation principles

NumErrors

greater_than

use_weights

Default: true

The proportion of predictions, for which the difference from the label value exceeds the specified value greater_than.

Calculation principles

SMAPE

use_weights

Default: true

Calculation principles
R2

use_weights

Default: true

Calculation principles

is the average label value:

MSLE

use_weights

Default: true

Calculation principles
MedianAbsoluteError
Calculation principles
Name Used for optimization User-defined parameters Formula and/or description
MAE +

use_weights

Default: true

Calculation principles
MAPE +

use_weights

Default: true

Calculation principles
Poisson +
Calculation principles
Quantile +

alpha

Default:  0.5

Calculation principles

RMSE +
Calculation principles
LogLinQuantile +

alpha

Default:  0.5

Calculation principles

Depends on the condition for the ratio of the label value and the resulting value:

Lq +

q

Default: Obligatory parameter

Calculation principles
Huber +
Calculation principles

Expectile +
Calculation principles

FairLoss
Calculation principles

NumErrors

greater_than

use_weights

Default: true

The proportion of predictions, for which the difference from the label value exceeds the specified value greater_than.

Calculation principles

SMAPE

use_weights

Default: true

Calculation principles
R2

use_weights

Default: true

Calculation principles

is the average label value:

MSLE

use_weights

Default: true

Calculation principles
MedianAbsoluteError
Calculation principles