I’d like to create a decision tree in python with a different split criterion than information gain, something like “1-information gain” (something like the opposite of impurity measure, like as similarity measure). Does already exist something like this? paper included. Thanks
Advertisement
Answer
Yes, it exists. There are many research papers:
https://pdfs.semanticscholar.org/5e44/d49b2268421d7ddf09d68be9aa689359b772.pdf