CreateSpaghettiVis: Create Spaghetti Visualizations

In google/glassbox: Glassbox package for RF visualization

Description

Creates "Spaghetti" visualization for random forest models to determine the true nature of the effectsCreate Spaghetti visualizations

Usage

CreateSpaghettiVis(rf.model, test.df, file, col1 = rgb(1, 0, 0, 1),
            col0 = rgb(0, 0, 0, 1), point.cex = 1, n.sample = 50,
            loess.span = .80,
            truncate.quantile.upper = .98, truncate.quantile.lower = .02)

Arguments

rf.model	fitted random forest model
test.df	hold out data set
file	output location

Examples

  library(glassbox)
  data(testdata)
  library(randomForest)
  rf.model <- randomForest(y ~ x + z, data=testdata)
  CreateSpaghettiVis(rf.model, testdata, "~/Google Drive/filename.pdf",
  col1=rgb(0, 0, 0, .05), n.sample = 1000)

  # Binary case
  testdata$w <- runif(1000) <= plogis(.5 * testdata$x - .7 * testdata$z)
  train <- testdata[1:800, ]
  test <- testdata[801:1000, ]
  rf.model.2 <- randomForest(as.factor(w) ~ x + z, train)
  CreateSpaghettiVis(rf.model.2, test, "~/Google Drive/filename2.pdf",
  col0=rgb(1, 0, 0, .05), col1=rgb(0, 0, 0, 1), n.sample=200)

google/glassbox documentation built on May 17, 2019, 7:47 a.m.