面對激烈競爭的環境,企業在有限的資源下,如何利用顧客歷史交易資料,透過資料探勘技術將顧客分群,找出潛藏有用的關聯與時序規則,協助企業在顧客關係管理的運作與實施,作為未來客戶經營發展策略的依據,為本研究想要探討的主要議題。 本研究利用個案公司三年來,透過網路的交易資料作為分析基準,運用最近購買日期、購買頻率、購買金額以及顧客關係長度等四個變數,透過分析層級程序法,以專家團體決策方法決定上述四個顧客購買行為變數之權重,最後依據加權後的顧客購買行為變數來衡量顧客終身價值。 利用此四個變數作為顧客分群的依據,以資料探勘技術期望值最大化演算法之集群分析,成功的將顧客分為「新進客戶」、「淺嘗即止客戶」、「潛力客戶」、「預警客戶」、「黏著客戶」及「忠誠客戶」等六群顧客群。 接著透過關聯法則Apriori演算法與時序集群分析演算法,分別找出各集群顧客購買產品的關聯及時序規則,最終透過規則的適用篩選,加上領域知識的探討,訂出各集群客製化交叉銷售及垂直銷售的行銷策略,供個案公司管理階層行銷決策之建議。
The main focus of this research is to discuss how industries can utilize basic information and transactional data of customers by applying data mining technique to further isolate those of who with hidden or inconspicuous data connections in order to achieve efficient customer relationship management. This study uses online transactions information for the past three years of the cast study company as samples for analysis. Recency (R), Frequency (F), Monetary (M) and Length (L) are the four variables chosen to be analyzed in the Analytic Hierarchy Process (AHP) and expert modeling to determine the weight and further evaluate Customer Lifetime Value (CLV) through those weighed variables. Using the above four variables as indexes for separating customers and by using data mining technique to analyze expectation- maximization algorithm, we can successfully segment customers into six groups- New Customers, Scratch the Surface Customers, Potential Customers, Cautious Customers, Adhesive Customers and Loyal Customers. Next, by applying Association Rules Apriori and Sequence Cluster Analysis we can determine purchase connections and sequences of each group. Finally through filters and domain know- how, we are able to customize cross and vertical sales marketing strategies which in turn provides valuable suggestions to the upper management marketing decisions.