classification - R package "tree": what is the difference between minsize and mincut? -
the tree
command constructing classification trees (using "tree" library) configured using tree.control
. tree.control
page explains minsize
, mincut
parameters follows:
mincut minimum number of observations include in either child node. weighted quantity; observational weights used compute ‘number’. default 5. minsize smallest allowed node size: weighted quantity. default 10.
to me, these 2 descriptions seem similar things. difference between mincut
, minsize
?
as understand it, classification, mincut
determines minimum number of observations required each class minsize
minimum number of observations required node.
for example suppose have 14 observations in node , deciding whether split. if 11 in class , 4 in class b shouldn't split because don't have @ least 5 of each class. if had 10 in class , 5 in class b split.
Comments
Post a Comment