새벽배송 식품 구매빈도 예측 및 상품 판매전략 연구

국문 초록 (Abstract)

최근 새벽배송 시장의 규모는 빠르게 성장하고 있다. 특히 코로나19 이후 그 규모는 2020년 2조원에서 2022년에 9조원대로 성장하였으며, 올해는 11조원대로의 성장이 전망되고 있다. 하지만 기업 간 경쟁은 갈수록 심화되고 있는 추세이기 때문에 새벽배송 기업에서는 차별화된 경쟁력 확보가 절실한 상황이다. 따라서 본 연구에서는 새벽배송 구매빈도에 영향을 미치는 중요한 변수를 확인하였으며, 새벽배송 기업에서 활용 가능한 머신러닝 분류모델을 제안하기 위하여 다양한 예측 알고리즘의 성능 비교를 통해 XG 부스트(XGBoost) 모델의 우수성을 발견하였다. 이후 상품 판매전략 도출을 위한 방법으로 먼저 장바구니 분석을 실시하여 총 3가지 상품군(간편식과 가공식품, 물과 가공식품, 건강기능식품과 가공식품)이 함께 구매하는 비율이 높은 것을 발견했다. 다음으로 이원배치 분산분석(Two-way ANOVA)을 통해서 새벽배송 구매빈도에 대한 연령대와 구매식품 사이의 상호작용 효과를 검증하였다. 이러한 결과들을 토대로 종합적인 데이터 분석을 실시하여 최종적으로 상품 판매전략 총 5가지(정기배송 서비스 전략, 온라인 배너 활용 묶음상품 추천 전략, 품질정보 제공의 다양화 전략, 타임세일 전략, 세대 맞춤형 트렌드 상품 추천 전략)를 제시하였다. 본 연구에서는 새벽배송 기업에서 활용 가능한 머신러닝 분류모델 제시를 통해 소비자의 구매빈도를 예측할 수 있게 하였으며, 데이터 분석 기법 및 상품 판매전략에 대한 시사점을 제공하였다.

다국어 초록 (Multilingual Abstract)

The scale of the early morning delivery market has been rapidly growing recently. Particularly, after the COVID-19 pandemic, the market size increased from 200 trillion won in 2020 to the 9 trillion won range in 2022, and it is expected to further grow to around 11 trillion won this year. However, due to intensifying competition among companies, securing differentiated competitiveness has become crucial for early morning delivery companies. Therefore, in this study, we identified significant variables influencing the purchase frequency of early morning delivery and discovered the superiority of the XGBoost model through a performance comparison of various prediction models, aiming to propose a machine learning classification model that can be utilized by early morning delivery companies. Subsequently, we conducted market basket analysis as a method for deriving product sales strategies and found that there is a high proportion of combined purchases in three product categories: home meal replacement and processed food, water and processed food, and health functional food and processed food. Furthermore, we verified the interaction effects between age groups and purchased food items on the purchase frequency of early morning delivery through two-way ANOVA. Based on these results, comprehensive data analysis was conducted, resulting in the proposal of five final product sales strategies: regular delivery service strategy, bundled product recommendation strategy utilizing online banners, strategy for diversifying quality information provision, time-sale strategy, and generation-specific trend product recommendation strategy. This study facilitates the prediction of consumer purchase frequency through the application of a machine learning classification model for early morning delivery companies, while also offering insights into data analysis techniques and sales strategies.

목차 (Table of Contents)

• 차 례
• 그림 차례 ······························································································ iv
• 표 차례 ································································································ v
• 국문 요약 ···························································································· vi
• 제 1장 서론 ···························································································· 1
• 1.1 연구 배경 및 목적 ············································································ 1
• 제 2장 이론적 배경 ··················································································· 2
• 2.1 온라인 상품 구매요인 관련 선행연구 ····················································· 3
• 2.2 새벽배송 식품 구매요인 관련 선행연구 ·················································· 3
• 제 3장 연구 방법 ····················································································· 6
• 3.1 연구 데이터 ··················································································· 7
• 3.2 데이터 전처리 ················································································· 9
• 3.3 장바구니 분석 ················································································ 10
• 3.4 머신러닝 분류 기법 ········································································· 11
• 3.4.1 서포트 벡터 머신 ······································································ 11
• 3.4.2 K-최근접 이웃 ········································································ 11
• 3.4.3 랜덤 포레스트 ········································································ 11
• 3.4.4 의사결정 나무 ········································································ 12
• 3.4.5 그래디언트 부스팅 ····································································· 12
• 3.4.6 XG 부스트 ········································································ 12
• 3.4.7 라이트 GBM ········································································ 13
• 제 4장 연구 결과 및 분석 ········································································· 13
• 4.1 변수 중요도 측정 및 변수 선택 ·························································· 13
• 4.2 데이터 분석 결과 ············································································ 14
• 4.2.1 인구통계학 정보 ········································································ 14
• 4.2.2 구매 시간대 ············································································· 16
• 4.2.3 구매식품 ················································································· 16
• 4.2.4 쇼핑몰 선택이유 ········································································ 17
• 4.2.5 구매이유 ················································································ 18
• 4.2.6 온라인식품 구매주기 ·································································· 19
• 4.3 장바구니 분석 결과 ········································································· 20
• 4.4 머신러닝 분류모델 성능 평가 ···························································· 21
• 4.5 이원배치 분산분석 결과 ···························································· 23
• 4.5.1 연령대와 건강기능식품의 분석 결과 ··············································· 23
• 4.5.2 연령대와 과일의 분석 결과 ·························································· 24
• 제 5장 상품 판매전략 ·············································································· 26
• 5.1 쇼핑몰 선택이유 및 장바구니 분석 기반 상품 판매전략 ··························· 27
• 5.1.1 정기배송 서비스 전략·································································· 29
• 5.1.2 온라인 배너 활용 묶음상품 추천 전략 ············································ 30
• 5.2 구매이유 및 이원배치 분산분석 기반 상품 판매전략 ································ 31
• 5.2.1 품질정보 제공의 다양화 전략 ······················································· 32
• 5.2.2 타임세일을 통한 상품판매 전략 ··················································· 33
• 5.2.3 세대 맞춤형 트렌드 상품 추천 전략 ··············································· 34
• 제 6장 결론 및 시사점 ············································································· 36
• 6.1 결론 ··························································································· 36
• 6.2 시사점 ························································································ 37
• 제 7장 한계점 및 향후연구 ······································································· 38
• 참고 문헌 ····························································································· 40
• Abstract ······························································································ 42