應用關鍵特徵值分析、分群和預測模型進行零售業顧客購買行為的縱向研究;Longitudinal Study of Retail Customer Purchase Behavior Using Feature Importance Analysis, Clustering, and Predictive Modeling

NCU Institutional Repository > 管理學院 > 企業管理研究所 > 博碩士論文 > Item 987654321/94951

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/94951

題名:	應用關鍵特徵值分析、分群和預測模型進行零售業顧客購買行為的縱向研究;Longitudinal Study of Retail Customer Purchase Behavior Using Feature Importance Analysis, Clustering, and Predictive Modeling
作者:	李結衣;Lee, Chieh-Yi
貢獻者:	企業管理學系
關鍵詞:	關鍵特徵值;顧客分群;顧客購買次數;RFM 模型;XGBoost模型;BG/NBD模型;Key Feature;Customer Segmentation;Purchase Frequency;RFM model;XGBoost model;BG/NBD model
日期:	2024-07-26
上傳時間:	2024-10-09 15:38:52 (UTC+8)
出版者:	國立中央大學
摘要:	了解顧客行為對於零售業的成功至關重要，在競爭激烈且瞬息萬變的市場環境中，企業需要深入洞察顧客需求和購買行為，以制定有效的行銷策略和提升顧客滿意度。顧客分群和顧客購買次數預測是了解顧客行為的兩種重要方法，然而過去針對零售業的研究中對於關鍵特徵值分析的應用較少，這是一個值得深入探討的領域，因為它能幫助企業辨識影響顧客購買行為的關鍵因素，並能夠進一步提升行銷策略的準確性和有效性。　　過去的研究未能充分整合關鍵特徵值分析、顧客分群和顧客購買次數預測這三個重要面向，因此本研究旨在整合這三種模型，並透過台灣某連鎖零售業的個案進行驗證。本研究之具體研究目標有三：(1) 使用XGBoost模型對顧客縱向資料進行關鍵特徵值分析，確認是否存在其他對顧客行為具有高度影響的變數；(2) 針對不同情境，採用兩階段分群法進行顧客分群，並了解各情境下的顧客特性與行為特徵；(3) 探討BG/NBD模型在零售業中的應用，預測顧客購買次數並評估其在實際應用中的表現。應用研究架構於台灣某連鎖零售業結果顯示，關鍵特徵值分析在所有情境下均得到相同結果，除RFM模型的三個變數（Recency、Frequency、Monetary）外，顧客的購買週期、促銷使用比例及類一產品購買比例也是影響顧客購買行為的重要變數；在顧客分群方面，各情境下的兩階段分群方法均有效地將顧客分群，並展示了其在提升分群品質和解釋力方面的優越性；在顧客購買次數預測方面，計算出的R-squared為 0.699，顯示預測模型具有良好的表現，能準確捕捉顧客購買次數的分佈趨勢。;Understanding customer behavior is crucial for the success of the retail industry. In a highly competitive and rapidly changing market environment, companies need deep insights into customer needs and purchasing behaviors to develop effective marketing strategies and enhance customer satisfaction. Customer segmentation and purchase frequency prediction are two essential methods for understanding customer behavior. However, past research on the retail industry has rarely applied key feature analysis, which is a worthwhile area for further exploration as it can help companies identify critical factors influencing customer purchasing behavior and further improve the accuracy and effectiveness of marketing strategies. Previous studies have not sufficiently integrated key feature analysis, customer segmentation, and purchase frequency prediction. Therefore, this study aims to integrate these three models and verify them through a case study of a chain retailer in Taiwan. The specific research objectives of this study are threefold: (1) to use the XGBoost model for key feature analysis on longitudinal customer data to identify other highly influential variables on customer behavior; (2) to conduct customer segmentation using a two-stage clustering method for different scenarios and understand customer characteristics and behavioral traits under each scenario; (3) to explore the application of the BG/NBD model in the retail industry to predict customer purchase frequency and evaluate its performance in practical applications. The application of the research framework to a chain retailer in Taiwan shows that key feature analysis consistently identifies the same critical variables across all scenarios. In addition to the three variables of the RFM model (Recency, Frequency, Monetary), customer purchasing cycle, promotion usage rate, and proportion of category 1 product purchases are also important variables influencing customer purchasing behavior. For customer segmentation, the two-stage clustering method effectively segmented customers in all scenarios, demonstrating its superiority in improving segmentation quality and explanatory power. For purchase frequency prediction, the calculated R-squared is 0.699, indicating that the prediction model performs well and accurately captures the distribution trend of customer purchase frequency.
顯示於類別:	[企業管理研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	49	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....