PatientLevelPrediction (version 4.3.10)

setGBMSurvival: Create setting for GBM Survival with python #' @description This creates a setting for fitting GBM surivial model. You need sksurv python install. To install this open your command line and type: conda install -c sebp scikit-survival

Description

Create setting for GBM Survival with python #' @description This creates a setting for fitting GBM surivial model. You need sksurv python install. To install this open your command line and type: conda install -c sebp scikit-survival

Usage

setGBMSurvival(
  loss = "coxph",
  learningRate = 0.1,
  nEstimators = c(100),
  criterion = "friedman_mse",
  minSamplesSplit = 2,
  minSamplesLeaf = 1,
  minWeightFractionLeaf = 0,
  maxDepth = c(3, 10, 17),
  minImpuritySplit = NULL,
  minImpurityDecrease = 0,
  maxFeatures = NULL,
  maxLeafNodes = NULL,
  presort = NULL,
  subsample = 1,
  dropoutRate = 0,
  seed = NULL,
  quiet = F
)

Arguments

loss

A string specifying the loss function to minimise (default: 'coxph' )

learningRate

A double specifying the learning rate (controls convergence speed)

nEstimators

An integer specifying how many trees to build

criterion

Default: 'friedman_mse'

minSamplesSplit

An integer specifying min samples per tree split (complexity)

minSamplesLeaf

An integer specifying min samples per leaf (complexity)

minWeightFractionLeaf

Lookup

maxDepth

An integer specifying the max depth of trees (complexity)

minImpuritySplit

A double or NULL specifying the minimum impurity split

minImpurityDecrease

will add

maxFeatures

will add

maxLeafNodes

will add

presort

will add

subsample

will add

dropoutRate

will add

seed

will add

quiet

will add

Details

Pick the hyper-parameters you want to do a grid search for

Examples

# NOT RUN {
gbmSurv <- setGBMSurvival(learningRate=c(0.1,0.01), nEstimators =c(10,50,100),
 maxDepth=c(4,10,17), seed = 2)
# }