약속 장소 추천을 ML 문제로 재정의하는 방법 - XGBoost 추천시스템 1편

Redefining Meetup Recommendation as ML Problem - XGBoost Recommender System Part 1

numpy>=2.1.0
pandas>=2.2.0
scikit-learn>=1.5.0
xgboost>=2.1.0
psycopg2-binary>=2.9.9

 # 사용자 행동 데이터를 종합하여 머신러닝 학습용 타깃 점수를 계산합니다.
def build_target_score(user_action):
    # 화면에 아예 노출되지 않은 장소는 학습할 의미가 없으므로 제외합니다.
    if user_action["is_exposed"] == 0:
        return None
    
    # 행동의 깊이에 따라 가중치를 다르게 부여합니다. (확정에 가까울수록 높은 점수)
    click_score = 0.2 if user_action["is_clicked"] == 1 else 0.0
    vote_score = 0.3 if user_action["is_voted"] == 1 else 0.0
    final_score = 0.4 if user_action["is_final_selected"] == 1 else 0.0
    
    # 5점 만점의 만족도 별점을 0에서 0.1 사이의 추가 보너스 점수로 변환합니다.
    rating = user_action.get("rating", 0)
    feedback_score = (rating / 5.0) * 0.1
    
 # 모든 점수를 합산하여 최종 점수를 반환합니다.
    return click_score + vote_score + final_score + feedback_score

{"room_id":"room_001","place_id":"place_a","is_exposed":1,"is_clicked":1,"is_voted":1,"is_final_selected":0,"rating":0}
{"room_id":"room_001","place_id":"place_b","is_exposed":1,"is_clicked":1,"is_voted":0,"is_final_selected":1,"rating":5}
{"room_id":"room_001","place_id":"place_c","is_exposed":1,"is_clicked":0,"is_voted":0,"is_final_selected":0,"rating":0}
{"room_id":"room_002","place_id":"place_d","is_exposed":0,"is_clicked":0,"is_voted":0,"is_final_selected":0,"rating":0}

#!/usr/bin/env bash
set -euo pipefail
 
python3 -m venv .venv-ml
source .venv-ml/bin/activate
 
python -m pip install --upgrade pip
pip install -r requirements-ml.txt
 
echo "Environment ready."

#!/usr/bin/env bash
set -euo pipefail
 
source .venv-ml/bin/activate
 
python ml/label_builder.py \
  --input ml/sample_actions.jsonl \
  --output ml/sample_actions_with_labels.jsonl

{"room_id": "room_001", "place_id": "place_a", "is_exposed": 1, "is_clicked": 1, "is_voted": 1, "is_final_selected": 0, "rating": 0, "target_score": 0.5}
{"room_id": "room_001", "place_id": "place_b", "is_exposed": 1, "is_clicked": 1, "is_voted": 0, "is_final_selected": 1, "rating": 5, "target_score": 0.7}
{"room_id": "room_001", "place_id": "place_c", "is_exposed": 1, "is_clicked": 0, "is_voted": 0, "is_final_selected": 0, "rating": 0, "target_score": 0.0}