In AlanInglis/vividOld: Variable Importance and Variable Interaction Displays

Step 1: install vivid

Use:

install.packages('vivid')

then

library(vivid)

Step 2: Create an mlr3 model:

For example, using simulated data from the Friedman benchmark problem 1, we create a mlr3 model and fit a ranger random forest First we load the required libraries:

library(mlr3)  # To create a model
library(mlr3learners)
library(ranger)

Then load the data:

myData <- genFriedman(noFeatures = 10, noSamples = 250, sigma = 1, bins = NULL, seed = NULL)

Then create the mlr3 model, using y as the response:

set.seed(1701)
fr_task  <- TaskRegr$new(id = "Friedman", backend = myData, target = "y")
set.seed(1701)
fr_lrn <- lrn("regr.ranger", importance = "permutation")
set.seed(1701)
fr_mod <- fr_lrn$train(fr_task)

Create a matrix of interaction/importance values using the vivid package:

set.seed(1701)
myMatrix <- vividMatrix(task = fr_task, model = fr_mod, gridSize = 30,                                normalize = FALSE, seed = NULL,
                         sqrt = FALSE, reorder = TRUE)

Create a HeatMap style plot displaying the importance on the diagonal and interactions on the off-diagonal:

plot(myMatrix, type = "heatMap", plotly = F,
          intLow = "floralwhite", intHigh = "dodgerblue4",
          impLow = "white", impHigh = "firebrick1", 
          minImp = NULL, maxImp = NULL, minInt = 0, maxInt = NULL)

Create a Network style plot displaying where the importance is displayed by both the size of the node (the bigger the node, the more important the variable) and the colour of the node (low to high values go from white to red)

plot(myMatrix, type = "network", thresholdValue = 0, label = FALSE,
            minInt = 0, maxInt = NULL, minImp = 0, maxImp=NULL,
            labelNudge = 0.05, layout = "circle")

Create a partial dependence pair style plot:

set.seed(1701)
ggpdpPairs(task = fr_task, model = fr_mod, gridsize = 10, mat = myMatrix)

Create a zenpath and zenplot layout of the partial dependence:

set.seed(1701)
zpath <- calcZpath(myMatrix, cutoff = 0.05)
pdpZenplot(task = fr_task, model = fr_mod, gridsize = 10)

Create a plot displaying all interactions:

plot(myMatrix, type = "allInteractions", top = 0)

Create a plot displaying the overall interaction strength:

interactionPlot(model =  fr_mod, data = myData, type = "barplot")

Create a plot displaying just the variable importance:

plot(myMatrix, type = "importance", plotType = "barplot")

Generate data from the Friedman benchmark problem 1:

myData <- genFriedman(noFeatures = 10, noSamples = 100, sigma = 1, bins = NULL, seed = NULL)
head(myData,3)

AlanInglis/vividOld documentation built on March 4, 2021, 12:44 a.m.