DecisionTree-class: Decision Trees for Random Forests

In npcooley/SynExtend: Tools for Working With Synteny Objects

Decision Trees for Random Forests

Description

DecisionTree objects comprising random forest models generated with RandForest.

Usage

## S3 method for class 'DecisionTree'
as.dendrogram(object, ...)

## S3 method for class 'DecisionTree'
plot(x, plotGain=FALSE, ...)

Arguments

object	an object of class DecisionTree to convert to class dendrogram.
x	an object of class DecisionTree to plot.
plotGain	logical; should the Gini gain or decrease in sum of squared error be plotted for each decision point of the tree?
...	For plot, further arguments passed to plot.dendrogram and text. Arguments prefixed with "text." (e.g., text.cex) will be passed to text, and all other arguments are passed to plot.dendrogram. For as.dendrogram, ... is further arguments for consistency with the generic definition.

Details

These methods help work with DecisionTree objects, which are returned as part of RandForest. Coercion to dendrogram objects creates a 'dendrogram' corresponding to the structure of the decision tree. Each internal node possesses the standard attributes present in a 'dendrogram' object, along with the following extra attributes:

• variable: which variable was used to split at this node.

• thresh: cutoff for partitioning points; values less than thresh are assigned to the left node, and those greater than to the right node.

• gain: change in the metric to maximize. For classification trees this is the Gini Gain, and for regression trees this is the decrease in sum of squared error.

Plotting allows for extra arguments to be passed to plot and text. Arguments prefixed with 'text' are passed to text, which controls the labeling of internal nodes. Common arguments used here are text.cex, text.adj, text.srt, and text.col. All other arguments are passed to plot.dendrogram. For example, col='blue' would change the dendrogram color to blue, whereas text.col='blue' would change the interior node labels to blue (but not the dendrogram itself).

Value

as.dendrogram returns an object of class 'dendrogram'. plot returns NULL invisibly.

Warning

These functions can be quite slow for large decision trees. Usage is discouraged for trees with more than 100 internal nodes.

Author(s)

Aidan Lakshman [e-mail]

See Also

RandForest

Examples

set.seed(199L)
n_samp <- 100L
AA <- rnorm(n_samp, mean=1, sd=5)
BB <- rnorm(n_samp, mean=2, sd=3)
CC <- rgamma(n_samp, shape=1, rate=2)
err <- rnorm(n_samp, sd=0.5)
y <- AA + BB + 2*CC + err

d <- data.frame(AA,BB,CC,y)
train_i <- 1:90
test_i <- 91:100
train_data <- d[train_i,]
test_data <- d[test_i,]

rf_regr <- RandForest(y~., data=train_data, rf.mode="regression", max_depth=5L)
if(interactive()){
  # Visualize one of the decision trees
  plot(rf_regr[[1]])
}

dend <- as.dendrogram(rf_regr[[1]])
plot(dend)

npcooley/SynExtend documentation built on Sept. 20, 2024, 11:58 a.m.