Interpretability of Decision Tree
# Tag:
- Source/KU_ML
Interpretability of Decision Tree
Decision Tree는 test function들의 conjuction으로 이루어져 있으므로, Deep Learning과 같은 Black Box 모델들과 달리 Interpretability가 높다.
- 비슷한 instance들 끼리 찾기 위한 fast localization으로 활용 가능하다.
- Feature selection에 사용할 수 있다.
data가 많아질수록 신뢰성이 높아지는 Rule Support를 가진다.
Feature Selection
Root Node에 가까워질수록 데이터를 판가름 할 수 있는 중요한 Attribute라는 의미를 지니게 된다.
즉, Root Node에서 부터 시작해 몇 개의 Attribute만 선택해 차원을 축소 가능하다.