제2회 코스포 x 데이콘 도서 추천 알고리즘 AI경진대회 채용

안녕하세요. 민만두우입니다.

전처리는 'Age' :이상치 제거 및 10살 단위 구간화, 'Year-Of-Publication' : 이상치 범주화,  'Location' : 텍스트 전처리 및 나라, 주, 도시로 분리 'location_country': 주와 도시 데이터로 가장 많이 함께 나온 나라로 대체 했습니다.

cbm.yaml 파일 예시

data:
    use_features: ["User-ID", "Book-ID", "Age","Year-Of-Publication","Publisher","location_country","Book-Author","location_state","location_city"]
    label: ["Book-Rating"]
    cat_features: ["User-ID", "Book-ID","Publisher","location_country","Book-Author","location_state","location_city"]
    preprocess_features: {"Age":"preprocess_age", "Year-Of-Publication":preprocess_year2}
    num_features: []
model:
    model_name: cbm_regressor
    cat_features: ["User-ID", "Book-ID","Age","Year-Of-Publication","Publisher","location_country","Book-Author","location_state","location_city"] #모델에 적용할 cat_features

train:
    seed: 42

CatBoostRegressor와 Optuna를 활용했고,  Optuna의 경우 맥북 M1 로컬에서 돌렸습니다.

감사합니다! :)