ensembler.regression {metaEnsembleR}

Ensemble Regressor Training & Prediction, Model Result Evaluation

Description

This function uses the base learners, and the final top layer learner to produce an ensemble prediction. The user can input the number of base learners and specify the final learner along with the train-validation-test data partition split ratio. The predictions on the unseen new data is the resultant of the ensemble meta-learning of the heterogeneous learners aimed to reduce the generalization error in the predictive models. Functions from caret are used for training and prediction of the base learners and the final learner.

Usage

ensembler.regression(data
  , outcomeVARIABLEINDEX
  , IndividualModels
  , TopLayerModel
  , dstr
  , dsv
  , dst
  , unseen_new_data
)

Arguments

data

Data to be used for training, validation, and test.

outcomeVARIABLEINDEX

Index of the response/outcome variable.

IndividualModels

Training of base learners.

TopLayerModel

Top layer final learner.

dstr

Training data split ratio.

dsv

Validation data split ratio.

dst

Testing data split ratio.

unseen_new_data

Prediction on unseen new data.

Value

data

Complete data or the indexed data.

outcomeVARIABLEINDEX

Integer value of the response/outcome variable index.

IndividualModels

A vector of base learners or standalone individual learner.

TopLayerModel

Name of the Top layer learner.

dstr

fraction integer. Not greater than 1. Total sum of dstr, dsv, dst should be equal to 1.

dsv

fraction integer. Not greater than 1. Total sum of dstr, dsv, dst should be equal to 1.

dst

fraction integer. Not greater than 1. Total sum of dstr, dsv, dst should be equal to 1.

unseen_new_data

Unseen new data provided either as a csv file or imported dataframe.

Note

The function ensembler.regression), returns a list with the following elements: ensembler_return[1]Test data predictions. ensembler_return[2]Prediction values. ensembler_return[3]Model Result. ensembler_return[4]Unseen data predictions.

Author(s)

Ajay Arunachalam

Examples

library("metaEnsembleR")
data("rock")
unseen_rock_data <- rock[30:48,]
ensembler_return <- ensembler.regression(rock[1:30,]
  , 4
     ,c('lm')
           , 'rf'
                , 0.40
                     , 0.30
                          , 0.30
                               , unseen_rock_data)
testpreddata <- data.frame(ensembler_return[1])


####Performance comparison#####
modelresult <- ensembler_return[3]
modelresult
#write.csv(modelresult[[1]], "performance_chart.csv")

####unseen data###
unseenpreddata <- data.frame(ensembler_return[4])