運用「基因K-Means演算法」劃分混合客服中心客戶群

現今競爭激烈的商業環境中，顧客服務中心已成為大多數企業與顧客們互動的主要部門。依據企業所制定的整體目標，顧客服務中心基本上需以最小的成本提供高品質的服務。舉例來說，某些混合型顧客服務中心，除了處理顧客交易業務的作業程序外，也必需提供售後服務解決方案。若顧客服務中心能將龐大且複雜的顧客資料有系統地建立顧客分群規則，將可使顧客服務中心內的資源有效地整合且最佳化。本研究個案是一家位於菲律賓馬卡蒂市（Makati City）的混合型顧客服務中心，以過去五年所有交易數據為基礎，顧客的終身價值依照LRFM （Length, Recency, Frequency, and Monetary）顧客關係模型設定為一個量化指標，而對於交易數據資料庫則使用CRISP（Cross Industry Standard Process）方法進行處理與應用。首先，數據資料庫進行標準化去除LRFM顧客關係模型中變數週期性的差異，經過基因與K-means演算法和CRISP方法分析後得到16個初始的群集，其群集的中心是由LFRM顧客關係模型中圖案分析得到，而各變數的權重是使用熵值法（Shannon Entropy）進行計算，最後透過合併相似LRFM顧客關係模型圖形得到9個顧客群集。本研究將對各個群集各別制訂適合其對應顧客群體的特定經營與銷售計劃，而本研究結果可提供混合型客戶服務中心一個長期的顧客關係管理解決方案。

關鍵字

顧客關係管理；顧客分群；顧客終身價值； CRISP方法；基因K-Means演算法；混合型客戶服務中心

並列摘要

Playing a crucial role in today’s competitive business environment, contact centers have been the primary source of customer interaction for most organizations. Driven by their common business goals to deliver a high quality service with minimum costs, centers such as blended contact centers act on operational basis in handling customer transactions. Optimization of resources will only be achieved if marketing decisions such as customer segmentation are appropriately established. This study proposes a practical way to segment customers of a blended contact center in Makati City, Philippines using a five-year transactional data. Customer lifetime value is assigned as a quantitative indicator based on the LRFM customer relationship model. This study then proceeds with the application of CRISP method, a powerful method used in implementing data-driven projects, on the transaction database. The database are normalized first to avoid periodic differences in the LRFM variables. 16 initial clusters have been obtained by using Genetic K-Means algorithm on data modeling and analysis phase of CRISP method. The cluster centroids are analyzed by the LFRM patterns, which the weight of each variables are calculated using Shannon Entropy. Finally, 9 customer segments have been achieved by merging clusters with similar LRFM patterns. Each segment is evaluated and classified to formulate a specific marketing plan suitable to its corresponding customer segment. The results of this research can help blended contact centers sustain a long term CRM with their customers.

並列關鍵字

Customer Relationship Management ； Customer Segmentation ； Customer Lifetime Value ； CRISP Method ； Genetic K-Means Algorithm ； Blended Contact Center

參考文獻

Chang, H.H. and Tsay, S.F. (2004). Integration of SOM and K-means in Data Mining Clus-tering: An Empirical Study of CRM and Profitability Evaluation. Journal of Infor-mation Management, 11 (4), 161-203.

Ahn, H.S., Righter, R., and Shanthikumar, J.G. (2005). Staffing Decisions for Heterogene-ous Workers with Turnover. Mathematical Methods of Operations Research, 62 (3), 499-514.

Aksin O.Z. and Harker, P.T. (1999). To Sell or Not to Sell: Determining the Tradeoffs be-tween Service and Sales in Retail Banking Phone Centers. Journal of Service Re-search, 2 (1), 19-33.

Aksin, O.Z. and Karaesmen, F. (2007). Characterizing the Performance of Process Flexibil-ity Structures. Operation Research Letters, 35 (4), 477-484.

Aksin, Z., Armony, M., and Mehrotra, V. (2007). The Modern Call Center: A Mul-ti-Disciplinary Perspective on Operations Management Research. Production and Operations Management, 16 (6), 665-688.

國際替代計量

運用「基因K-Means演算法」劃分混合客服中心客戶群

未授權

主題瀏覽