Visualize Recursive Partitioning and Regression Trees (rpart object)

Visualize Recursive Partitioning and Regression Trees rpart. Have a look to visTreeEditor to edity and get back network, or to visTreeModuleServer to use custom tree module in R

visTree(
  object,
  data = NULL,
  tooltipColumns = if (!is.null(data)) {     1:ncol(data) } else {     NULL },
  main = "",
  submain = "",
  footer = "",
  direction = "UD",
  fallenLeaves = FALSE,
  rules = TRUE,
  simplifyRules = TRUE,
  shapeVar = "dot",
  shapeY = "square",
  colorVar = NULL,
  colorY = NULL,
  colorEdges = "#8181F7",
  nodesFontSize = 16,
  edgesFontSize = 14,
  edgesFontAlign = "horizontal",
  legend = TRUE,
  legendNodesSize = 22,
  legendFontSize = 16,
  legendWidth = 0.1,
  legendNcol = 1,
  legendPosition = "left",
  nodesPopSize = FALSE,
  minNodeSize = 15,
  maxNodeSize = 30,
  highlightNearest = list(enabled = TRUE, degree = list(from = 50000, to = 0), hover =
    FALSE, algorithm = "hierarchical"),
  collapse = list(enabled = TRUE, fit = TRUE, resetHighlight = TRUE, clusterOptions =
    list(fixed = TRUE, physics = FALSE)),
  updateShape = TRUE,
  tooltipDelay = 500,
  digits = 3,
  height = "600px",
  width = "100%",
  export = TRUE
)

Arguments

object

rpart, rpart object

data

data.frame, adding mini-graphics in tooltips using sparkline and tooltipColumns ?

tooltipColumns

numeric, indice of columns used in tooltip. All by default. So, we add boxplot / pie focus on sub-population vs all population using sparkline package. NULL to disable.

main

Title. See visNetwork

submain

Subtitle. See visNetwork

footer

Footer. See visNetwork

direction

character, The direction of the hierarchical layout. The available options are: UD, DU, LR, RL. To simplify: up-down, down-up, left-right, right-left. Default UD. See visHierarchicalLayout

fallenLeaves

boolean leaf nodes at the bottom of the graph ? Default to FALSE

rules

boolean, add rules in tooltips ? Default to TRUE

simplifyRules

boolean, simplify rules writing

shapeVar

character, shape for variables nodes See visNodes

shapeY

character, shape for terminal nodes See visNodes

colorVar

character, colors to use or data.frame To set color of variables. 2 columns :

"variable" : names of variables
"color" : colors (in hexa). See examples

colorY

if classification tree : character colors to use or data.frame 2 columns :

"modality" : levels of Y
"color" : colors (in hexa)

if regression tree : character, 2 colors (min and max, in hexa)

colorEdges

character, color of edges, in hexa. Default to #8181F7

nodesFontSize

numeric, size of labels of nodes. Default to 16

edgesFontSize

numeric, size of labels of edges Default to 14

edgesFontAlign

character, for edges only. Default tp 'horizontal'. Possible options: 'horizontal' (Default),'top','middle','bottom'. See visEdges

legend

boolean, add legend ? Default TRUE. visLegend

legendNodesSize

numeric, size of nodes in legend. Default to 22

legendFontSize

numeric, size of labels of nodes in legend. Default to 16

legendWidth

numeric, legend width, between 0 and 1. Default 0.1

legendNcol

numeric, number of columns in legend. Default 1

legendPosition

character, one of "left" (Default) or "right"

nodesPopSize

boolean, nodes sizes depends on population ? Default to FALSE

minNodeSize

numeric, in case of nodesPopSize, minimum size of a node. Default to 15. Else, nodes size is minNodeSize + maxNodeSize / 2

maxNodeSize

numeric, in case of nodesPopSize, maximum size of a node. Default to 30. Else, nodes size is minNodeSize + maxNodeSize / 2

highlightNearest

list, Highlight nearest nodes. See visOptions

collapse

list, collapse or not using double click on a node ? See visOptions

updateShape

boolean, in case of collapse, udpate cluster node shape as terminal node ? Default to TRUE

tooltipDelay

numeric, delay for tooltips in millisecond. Default 500

digits

numeric, number of digits. Default to 3

height

character, default to "600px"

width

character, default to "100%"

export

boolean, add export button. Default to TRUE

Value

a visNetwork object

References

See online documentation https://datastorm-open.github.io/visNetwork/

Examples


if (FALSE) { # \dontrun{

library(rpart)

# Basic classification tree
res <- rpart(Species~., data=iris)
visTree(res, data = iris, main = "Iris classification Tree")

# Basic regression tree
res <- rpart(Petal.Length~., data=iris)
visTree(res, edgesFontSize = 14, nodesFontSize = 16)

# Complex tree
data("solder")
res <- rpart(Opening~., data = solder, control = rpart.control(cp = 0.00005))
visTree(res, data = solder, nodesPopSize = TRUE, minNodeSize = 10, 
  maxNodeSize = 30, height = "800px")

# ----- Options
res <- rpart(Opening~., data = solder, control = rpart.control(cp = 0.005))

# fallen leaves + align edges label & size
visTree(res, fallenLeaves = TRUE, height = "500px", 
 edgesFontAlign = "middle", edgesFontSize = 20)

# disable rules in tooltip, and render tooltip faster
# enable hover highlight
visTree(res, rules = FALSE, tooltipDelay = 0, 
 highlightNearest = list(enabled = TRUE, degree = list(from = 50000, to = 0), 
 hover = TRUE, algorithm = "hierarchical"))

# Change color with data.frame
colorVar <- data.frame(variable = names(solder), 
 color = c("#339933", "#b30000","#4747d1","#88cc00", "#9900ff","#247856"))
 
colorY <- data.frame(modality = unique(solder$Opening), 
 color = c("#AA00AA", "#CDAD15", "#213478"))

visTree(res, colorEdges = "#000099", colorVar = colorVar, colorY = colorY)

# Change color with vector
visTree(res, colorEdges = "#000099", 
    colorVar = substring(rainbow(6), 1, 7), 
    colorY = c("blue", "green", "orange"))
    
    
 # Use visNetwork functions to add more options
visTree(res) %>% 
    visOptions(highlightNearest = TRUE)


} # }

Visualize Recursive Partitioning and Regression Trees (rpart object)

Arguments

Value

References

See also

Examples