運用資料探勘技術探討顧客價值與消費行為之研究─以零售業連鎖專賣店為例

基於顧客的價值有高低，顧客的行為各不同，國內零售業連鎖專賣店，為了加強顧客管理並使交易順利，近年來紛紛採用會員制度，也因此會員顧客的交易多能獲得完整的紀錄。本研究於企業資源有限前提下，為避免行銷預算被浪費與錯用，欲利用資料庫中個別顧客的購買記錄，來找出不同顧客價值之顧客群，並提出相對應的顧客經營策略。本研究以國內零售業某居家用品專賣店所定義「活躍性消費顧客」為研究對象。研究方法首先採用RFML變數（基於RFM模型）透過Kohonen網路與K-Means兩階段的集群分析，成功的將顧客分為「核心顧客」、「潛力顧客」、「預警顧客」、「新進顧客」及「消耗資源顧客」等五種價值顧客群。其次，對於顧客消費行為之探討，則透過決策樹C5.0、類神經網路與邏輯斯迴歸等演算法，依據顧客價值分群結果，分別建立顧客價值分群預測模型、顧客檔期活動回應預測模型與商品類別消費偏好評分模型，以應用於擬定顧客經營策略與資料庫行銷活動的規畫。相較於以單一變數消費金額進行顧客區隔之傳統行銷區隔模式，本研究利用資料探勘技術之間的互補性，透過研究實證發展出準確度與可用性均佳的各類主題式分析模型與規則，可協助企業發展深度的顧客洞察能力，進而逐步建立行銷專屬顧客知識庫，以創造最大的收益並達到行銷活動投資報酬率最佳化。

關鍵字

資料探勘； RFML顧客價值；集群分析；決策樹；類神經網路；邏輯斯迴歸

並列摘要

Base on the different values upon the customer categories and varieties of the customer behaviors, the national-wide retail stores tend to adopt the membership system in order to enhance the customer relationship management and business deals. In that case, the transaction of the membership customers was recorded completed. Even with the limitation upon the enterprise resources, this study not only intend to define the customer groups in different values as using the individual customer's purchasing record from the data base, but also bring up the corresponding management strategies, to avoid the misused and waste upon the marketing budget. The study uses national retails home furnishing stores "active customer" as an empirical case. The study (based on RFML four variables) design used Kohonen Net & K-Means two stages clustering successfully segment active customer to five groups - core customers, potential customers, cautious customers, new customers and spend resource customers. And then the study apply C5.0、Neural Network & Logistic Regression algorithm on the result of clustering to establish customer value and behavior prediction model, besides the forecast customer models of the activities respond and consumer preferences categories scoring model. All these are expected to apply on the development of customer business strategy and the planning upon the database marketing activities. Compared to a single variable amount of consumption for customer segmentation model of the traditional marketing segmentation, this research has developed the capable and precise models and regulations of kinds of categories in thematic analysis by using the complementary from the data mining systems and classification model created upon this study. These models can assist the enterprises develop a deeper relationship with the customers and in further step to precisely achieve the target consuming behaviors for database marketing project, for creating the biggest benefit achievement also the best result on the return of investment upon the marketing activities.

並列關鍵字

Data Mining ； RFML Customer Value ； Cluster Analysis ； Decision Tree ； Neural Network ； Logistic Regression.

參考文獻

張心馨、蔡憲富（2004）。以Data Mining 技術結合SOM 和K-Mean 的消費者分群方法於顧客關係管理和績效獲利性評估。資訊管理學報，11(4)，161-203。

張語恬、朱基銘、周雨青、楊燦、盧瑜芬、白健佑、…羅慶徽（2007）。比較三種資料探勘演算法預測子宮頸癌五年存活的外部通用性效能。臺灣家庭醫學雜誌。17(4)，222-238。

吳岱壎（2012）。運用資料採礦技術於零售業之研究—以居家用品專賣店為例。台北市：國立臺北大學企業管理研究所碩士論文。

劉復苓、邱天欣（譯）（2003）。CRM關鍵32堂課（原作者：J. Freeland）。台北市：美商麥格羅．希爾國際股份有限公司。（原著出版年：2003）

Berger, P. D., & Nasr, N. I. (1998). Customer Lifetime value: Marketing Models and Applications. Journal of Interactive Marketing, 12(1), 17-30.

國際替代計量

運用資料探勘技術探討顧客價值與消費行為之研究─以零售業連鎖專賣店為例

未授權

主題瀏覽