plotProximity: Plot Random Forest Proximity Scores

In rfPermute: Estimate Permutation p-Values for Random Forest Importance Metrics

Plot Random Forest Proximity Scores

Description

Create a plot of Random Forest proximity scores using multi-dimensional scaling.

Usage

plotProximity(
  x,
  dim.x = 1,
  dim.y = 2,
  class.cols = NULL,
  legend.type = c("legend", "label", "none"),
  legend.loc = c("top", "bottom", "left", "right"),
  point.size = 2,
  circle.size = 8,
  circle.border = 1,
  group.type = c("ellipse", "hull", "contour", "none"),
  group.alpha = 0.3,
  ellipse.level = 0.95,
  n.contour.grid = 100,
  label.size = 4,
  label.alpha = 0.7,
  plot = TRUE
)

Arguments

x	a rfPermute or randomForest model object.
dim.x, dim.y	numeric values giving x and y dimensions to plot from multidimensional scaling of proximity scores.
class.cols	vector of colors to use for each class.
legend.type	type of legend to use to label classes.
legend.loc	character keyword specifying location of legend. Can be "bottom", "top", "left", "right".
point.size	size of central points. Set to NULL for no points.
circle.size	size of circles around points indicating classification. Set to NULL for no circles.
circle.border	width of circle border.
group.type	type of grouping to display. Ignored for regression models.
group.alpha	value giving alpha transparency level for group shading. Setting to 0 produces no shading.
ellipse.level	the confidence level at which to draw the ellipse.
n.contour.grid	number of grid points for contour lines.
label.size	size of label if legend.type = `label`.
label.alpha	transparency of label background.
plot	logical determining whether or not to show plot.

Details

Produces a scatter plot of proximity scores for dim.x and dim.y dimensions from a multidimensional scale (MDS) conversion of proximity scores from a randomForest object. For classification models, points are colored according to original (inner) and predicted (outer) class.

Value

a list with:

prox.mds	the MDS scores of the selected dimensions
g	ggplot object

Author(s)

Eric Archer [e-mail]

Examples

library(randomForest)
data(symb.metab)

rf <- randomForest(type ~ ., symb.metab, proximity = TRUE)

# With confidence ellipses
plotProximity(rf)

# With convex hulls
plotProximity(rf, group.type = "hull")

# With contours
plotProximity(rf, group.type = "contour")

# Remove the points and just show ellipses
plotProximity(rf, point.size = NULL, circle.size = NULL, group.alpha = 0.5)

# Labels instead of a legend
plotProximity(rf, legend.type = "label", point.size = NULL, circle.size = NULL, group.alpha = 0.5)

rfPermute documentation built on Aug. 24, 2023, 1:08 a.m.