Issue
I try to understand, how the algoritm select feature for the root node. Which criterion and thresholds are used for this?
Solution
According to the documentation for DecisionTreeClassifier in scikit-learn (using the default arguments) the algorithm will consider every possible feature and every possible split on that feature and the split with the greatest quality according to the criterion
measure, (in other words: the split that produces the best separation of all the classes) will be chosen first and placed into the root node. More explanation on your specific problem can be obtained by following this scikit-learn tool.
Answered By - TC Arlen
0 comments:
Post a Comment
Note: Only a member of this blog may post a comment.