XGBMetricForOrderedLogit: Custom Objective/Evaluate Functions Class for XGBoost
In atamakita/xgboostOLR: Custom Objective/Evaluate Functions for xgboost
Description
Details
Public fields
Active bindings
Methods
Note
Description
Custom Objective/Evaluate Functions Class for XGBoost
Custom Objective/Evaluate Functions Class for XGBoost
Details
R6 Class defining custom Objective/Evaluate for XGBoost
Public fields
Criterionvalues between each labels for Ordered Logit Model
notchdiff_evalwidth of notch when evaluating hit ratio
Active bindings
Criterionvalues between each labels for Ordered Logit Model
notchdiff_evalwidth of notch when evaluating hit ratio
Methods
Public methods
-
-
-
-
-
-
-
-
-
-
Method set_criterion()
Set Criterions between each labels for Ordered Logit Model
Usage
XGBMetricForOrderedLogit$set_criterion(x)
Arguments
xA double vector
Method return_criterion()
Return criterions between each labels for Ordered Logit Model
Usage
XGBMetricForOrderedLogit$return_criterion()
Returns
Criterion set in this class
Method set_notchdiff_eval()
Set notch width when evaluating hit ratio by eval_hitratio()
Usage
XGBMetricForOrderedLogit$set_notchdiff_eval(x)
Arguments
xAn integer, notch width when evaluate hit ratio
Method return_notchdiff_eval()
Return notch width
Usage
XGBMetricForOrderedLogit$return_notchdiff_eval()
Returns
private$notchdiff_eval
Method pred_class_criterion()
Function to predict labels comparing preds and criterions (not maximum probability)
Usage
XGBMetricForOrderedLogit$pred_class_criterion(preds)
Arguments
predsA double vector, margin score from xgboost before logistic transformation
Returns
Integer label, if [-Inf, criterion[1]) label is 1, else if [criterion[1], criterion[2]), label is 2, ...
Method pred_class_maxprob()
Function to predict labels at maximum probability
Usage
XGBMetricForOrderedLogit$pred_class_maxprob(preds)
Arguments
predsA double vector, margin score from xgboost before logistic transformation
Returns
Integer label at max probability
Method obj_ordered_logit()
Custom Objective Function of Ordered Logit Model
https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$obj_ordered_logit(preds, dtrain)
Arguments
predsA double vector, margin score at t-1
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector).
$grad gradient of loss function.
$hess hessian of loss function.
Method eval_logloss()
Custom Evaluate Function of Ordered Logit Model (logloss)
https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$eval_logloss(preds, dtrain)
Arguments
predsA double vector, margin score
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector).
$metric name of metric.
$value value of metric.
Method eval_hitratio()
Custom Evaluate Function of Ordered Logit Model (hit ratio)
https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$eval_hitratio(preds, dtrain)
Arguments
predsA double vector, margin score
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector).
$metric name of metric.
$value value of metric.
Method clone()
The objects of this class are cloneable with this method.
Usage
XGBMetricForOrderedLogit$clone(deep = FALSE)
Arguments
deepWhether to make a deep clone.
Note
Add -Inf/Inf both side of input x
x is sorted ascending
criterion[1] = -Inf,
criterion[2] = x[1],
...
criterion[nClass] = x[nClass - 1],
criterion[nClass + 1] = Inf
x = 1 -> hit ratio is calculated by |act - est| < x / length(act)
grad, hess were referred to URL below.
https://www.slideshare.net/TakujiTahara/201200229-lt-dsb2019-ordered-logit-model-for-qwk-tawara#23
self$notchdiff_eval is set by set_notchdiff_eval
atamakita/xgboostOLR documentation built on Aug. 8, 2020, 5:39 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Description Details Public fields Active bindings Methods Note
Description
Custom Objective/Evaluate Functions Class for XGBoost
Custom Objective/Evaluate Functions Class for XGBoost
Details
R6 Class defining custom Objective/Evaluate for XGBoost
Public fields
Criterionvalues between each labels for Ordered Logit Model
notchdiff_evalwidth of notch when evaluating hit ratio
Active bindings
Criterionvalues between each labels for Ordered Logit Model
notchdiff_evalwidth of notch when evaluating hit ratio
Methods
Public methods
Method set_criterion()
Set Criterions between each labels for Ordered Logit Model
Usage
XGBMetricForOrderedLogit$set_criterion(x)
Arguments
xA double vector
Method return_criterion()
Return criterions between each labels for Ordered Logit Model
Usage
XGBMetricForOrderedLogit$return_criterion()
Returns
Criterion set in this class
Method set_notchdiff_eval()
Set notch width when evaluating hit ratio by eval_hitratio()
Usage
XGBMetricForOrderedLogit$set_notchdiff_eval(x)
Arguments
xAn integer, notch width when evaluate hit ratio
Method return_notchdiff_eval()
Return notch width
Usage
XGBMetricForOrderedLogit$return_notchdiff_eval()
Returns
private$notchdiff_eval
Method pred_class_criterion()
Function to predict labels comparing preds and criterions (not maximum probability)
Usage
XGBMetricForOrderedLogit$pred_class_criterion(preds)
Arguments
predsA double vector, margin score from xgboost before logistic transformation
Returns
Integer label, if [-Inf, criterion[1]) label is 1, else if [criterion[1], criterion[2]), label is 2, ...
Method pred_class_maxprob()
Function to predict labels at maximum probability
Usage
XGBMetricForOrderedLogit$pred_class_maxprob(preds)
Arguments
predsA double vector, margin score from xgboost before logistic transformation
Returns
Integer label at max probability
Method obj_ordered_logit()
Custom Objective Function of Ordered Logit Model https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$obj_ordered_logit(preds, dtrain)
Arguments
predsA double vector, margin score at t-1
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector). $grad gradient of loss function. $hess hessian of loss function.
Method eval_logloss()
Custom Evaluate Function of Ordered Logit Model (logloss) https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$eval_logloss(preds, dtrain)
Arguments
predsA double vector, margin score
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector). $metric name of metric. $value value of metric.
Method eval_hitratio()
Custom Evaluate Function of Ordered Logit Model (hit ratio) https://github.com/dmlc/xgboost/blob/master/R-package/demo/custom_objective.R
Usage
XGBMetricForOrderedLogit$eval_hitratio(preds, dtrain)
Arguments
predsA double vector, margin score
dtrainA xgb.DMatrix created by xgboost::xgb.DMatrix
Returns
A named list(both double vector). $metric name of metric. $value value of metric.
Method clone()
The objects of this class are cloneable with this method.
Usage
XGBMetricForOrderedLogit$clone(deep = FALSE)
Arguments
deepWhether to make a deep clone.
Note
Add -Inf/Inf both side of input x
x is sorted ascending
criterion[1] = -Inf,
criterion[2] = x[1],
...
criterion[nClass] = x[nClass - 1],
criterion[nClass + 1] = Inf
x = 1 -> hit ratio is calculated by |act - est| < x / length(act)
grad, hess were referred to URL below. https://www.slideshare.net/TakujiTahara/201200229-lt-dsb2019-ordered-logit-model-for-qwk-tawara#23
self$notchdiff_eval is set by set_notchdiff_eval
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.