Prediction Object for Ordinal Regression

Source: R/PredictionOrdinal.R

This object stores the predictions returned by a learner of class LearnerOrdinal.

Format

R6Class object inheriting from Prediction.

Usage

Inherits from Prediction

# Construction
p = PredictionOrdinal$new(task, response, prob)

# Members
p$predict_types
p$row_ids
p$truth

# S3 methods
as.data.table(p)

Arguments

  • task (Task): Task used for prediction. Used to extract row_ids and truth. Set to NULL to skip all argument checks during initialization. Slots p$row_ids and p$truth need to be set manually in this case.

  • response (factor() | ordered()): Vector of predicted class labels.

  • prob (matrix): Numeric matrix of rank probabilities with one column for each rank in task$class_names and one row for each observation in the test set.

Details

  • $new() initializes a new object of class Prediction.

  • $predict_types (character) stores the predict types available: a subset of c("response", "prob").

  • $response stores the predicted values.

  • row_ids stores the row IDs.

  • $truth stores the true rank vector.

  • The prediction object can be transformed to a simple data.table() with data.table::as.data.table.

Examples

library(mlr3)
task = mlr_tasks$get("winerating")
learner = mlr_learners$get("ordinal.clm")
p = learner$train(task)$predict(task)

#> Warning: (1) Hessian is numerically singular: parameters are not uniquely determined 
#> In addition: Absolute convergence criterion was met, but relative criterion was not met
#> Warning: predictions from column rank-deficient fit may be misleading
head(as.data.table(p))

#>    row_id truth response
#> 1:      1     2        2
#> 2:      2     3        3
#> 3:      3     3        3
#> 4:      4     4        4
#> 5:      5     4        4
#> 6:      6     4        4