•
모델: Random Forest Regressor, Catboost(시간 여유 될 때)
•
XAI 기법: LIME, SHAP
•
실험 절차:
1.
Data preparation → Training dataset
2.
Train model and apply XAI techniques → LIME feature importance / SHAP values
3.
Summarize XAI results → Ranking results of features
4.
Experiments
a.
Feature selection (recursive elimination)
b.
Performance evaluation for feature subset (training time, loss)
•
코드 구성:
◦
data_preparation.py
◦
train_and_xai.py
◦
performance_evaluation.py
References
1.
•
Random Forest Regressor 모델 매개변수 설정 방법
2.
•
모델 저장 및 불러오기 방법
3.
4.
•
SHAP 코드 예시
5.
•
Random Forest Regressor + SHAP 코드, impurity
