forestFeatures: Extract Vectors of Ranked and Selected Features From a Random...

In ClassifyR: A framework for cross-validated classification problems, with applications to differential variability and differential distribution testing

Description

Provides a ranking of features based on the total decrease in node impurities from splitting on the variable, averaged over all trees. Also provides the selected features which are those that were used in at least one tree of the forest.

Usage

  ## S4 method for signature 'randomForest'
forestFeatures(forest)

Arguments

forest

A trained random forest which was created by randomForest.

Value

An list object. The first element is a vector or data frame of features, ranked from best to worst using the Gini index. The second element is a vector or data frame of features used in at least one tree.

Author(s)

Dario Strbenac

Examples

      if(require(randomForest))
      {
        genesMatrix <- sapply(1:25, function(sample) c(rnorm(100, 9, 2)))
        genesMatrix <- cbind(genesMatrix, sapply(1:25, function(sample)
                                          c(rnorm(75, 9, 2), rnorm(25, 14, 2))))
        classes <- factor(rep(c("Poor", "Good"), each = 25))
        colnames(genesMatrix) <- paste("Sample", 1:ncol(genesMatrix))
        rownames(genesMatrix) <- paste("Gene", 1:nrow(genesMatrix))
        trainingSamples <- c(1:20, 26:45)
        testingSamples <- c(21:25, 46:50)
    
        trained <- randomForestTrainInterface(genesMatrix[, trainingSamples],
                                              classes[trainingSamples], ntree = 10)
                                        
        forestFeatures(trained)
      }

ClassifyR documentation built on Nov. 8, 2020, 6:53 p.m.