| gini {lazy.tree} | R Documentation |
gini, entropy, mis.match are for categorical criterion and
variance is for numerical criterion.
Japanese help file: gini_JPH
gini(y)
centropy(y)
mis.match(y)
entropy(y)
variance(y)
y |
The input vector (numeric) |
These impurity measures are not weighted by the # of observations in the set.
gini Gini impurity measure
entropy cross entropy impurity measure
mis.match proportion of observations whose predicted value
is not the same as the mode,
variance variance of the criterion in the subset.
This is equal to the resudial sum of squares divided by n.
entropy can be interpreted as minus the log likelihood of
the multinomial distribution.