Decision trees & continuous-valued attributes -
how decision tree handle continuous valued attributes? know attributes such sex
have 2 arcs denoting male
or female
. how continuous valued attributes such age
? not possible create arcs each possible value? create age-range sets? example having 3 sub-branches labeled <18
, <30
, >30
? if how decide on creation of these age-groups?
some algorithms cart evaluates possible splits using gini index or other impurity functions. sort attributes , @ impurity of each split. drawback of method is greedy, need @ every possible split. see image here.
Comments
Post a Comment