SuffStats: Create a SuffStats Object
In ZJUguquan/OnlineRandomForest: The implemention of Online Random Forest.

Description Usage Arguments Format Details Value Fields Methods Author(s) Examples

SuffStats is a class of R6. It has provide some important statistics(fields or functions) inside the element of every (leaf) node in an ORT object.

1	SuffStats$new(numClasses = 0)

numClasses

A nonnegative integer indicates how many classes when solve a classifation problem. Default 0 for regression. If numClasses > 0, then do classifation.

R6Class object.

Note that SuffStats may only be seen in leaf nodes of an ORT tree. See details in description of each field or method.

Object of R6Class, Object of SuffStats.

node.n: Number of samples under current node. Default 0.
classify: TRUE for classification and FALSE for Regression.
node.counts: If classification, counts of each y value in current node.
node.sum: Sum of the y value in current node.
node.square.sum: Sum of the y's square in current node.

update(y)

When a sample comes in current node, update ORT with the sample's y value.

y - The y value of a sample.

reset()

Reset all fields to NULL. Currently not used.

isLeaf()

TRUE if current node is a leaf node otherwise FALSE.

pred()

Return the prediction of current leaf node. Will return an integer for classification or an numeric for regression.

impurity()

Return the impurity of current node. Computed via information entropy.

Quan Gu

# classification
sa <- SuffStats$new(numClasses = 3)
sa$update(2);sa$update(1);sa$update(2)
sa$pred()
sa$node.counts
sa$node.n
sa$impurity()

# Regression
sb <- SuffStats$new(numClasses = 0)
sb$update(1);sb$update(2);sb$update(3)
sb$pred()
sb$node.sum;sb$node.square.sum
sb$impurity()