There is a parameter named minnumobj in the options of the j48 tree algorithm in weka. This parameter indicates the minimum number of participants to be in a leaf. Does this parameter have an optimum value regarding the instances or attributes? Or should I set the value that achieves the highest classification success? For example, I have 6500 instances and 9 attributes and I got highest success when minnumobj=100.
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
If there were an optimum in terms of number of instances or attributes, then that would be set in the program. No, such a parameter generally has an optimum that depends on the data itself, and needs to be tuned to get optimal performance.
answered Dec 29, 2020 at 19:09