應用資料探勘於顧客分群之研究－以製藥業為例

USING DATA MINING TECHNIQUE FOR CUSTOMERS CLUSTERING – AN EXAMPLE OF PHARMACEUTICAL MANUFACTURERS

曹毓齡(Yu-Ling Tsao)

指導教授 ： 陳煇煌

大同大學/經營設計學院/資訊經營學系所/碩士(2009年)

摘要

根據行政院衛生署最新統計結果，國內製藥廠高達244家，從IMS資料顯示，2008年我國1,187億元的藥品市場中，本國藥廠的市佔率僅22%，在台灣市場狹小，市場競爭卻非常激烈，且研發創新薄弱、研發人才斷層、新藥研發亦不是短時間可達到、美國FDA態度保守，使得新藥核准速度趨緩、智慧財產權的挑戰、再加上日趨嚴格的法規審核、原料藥持續成長，近年來金融風暴及健保藥價大幅調降的雙重打擊，使得營收和獲利受創，製藥業的經營備受威脅。
　　從相關文獻探討中，指出顧客價值，是在一段期間內顧客對於企業的潛在貢獻，本研究個案公司為製藥業，其顧客為醫院、診所及藥局的用藥人員，均屬於專業技術人員，病患通常會接受專業醫師或藥師的建議去選用適合的藥品，因此製藥業若想在競爭多變的環境中脫穎而出，與醫師或藥師建立良好的顧客關係，創造顧客價值，顯得相當重要。
本研究目的，期望能從客戶資料庫中發掘其交易特徵，將顧客分群，進而針對各群的顧客提出行銷建議，及適當的管理意涵，期許管理者能更妥善的運用有限資源，並協助管理者決策參考。
　　基於以上，本研究以國內製藥公司為樣本，進行顧客價值分析，將客戶資料及銷售資料，運用顧客價值分析(RFM)及K平均數方法(K-mean method)，將顧客予以分群，其輸入變數為RFM、最近購買日(R)、購買頻率(F)及平均購買金額(M)及總購買金額(TM)，並將兩種方法的結果做比較，研究結果RFM分析方法，將顧客分成13群，由數據顯示發現各分群間特徵不明顯且相似度極高，另一採用SQL Server 2005 Analysis Services (SSAS)分析，將顧客分為10群，採用10群亦為系統預設值，可以避免分太多群，造成資源分散，研究也發現K-mean模式分析的效果較好，較能看出各群間不同的交易特徵，最後，歸納出10種不同投資行為之群集，做為該群集之命名，再針對不同群集客戶特性來擬訂不同行銷策略，期望能提供實資上的管理決策之參考。

關鍵字

顧客價值(RFM) ； 美國食品藥物管理局(FDA) ； 顧客關係管理(CRM) ； 資料探勘 ； 群集分析

並列摘要

According to the National Department of Health, the latest statistics, the domestic pharmaceutical up to 244, from the IMS data show that in 2008 Taiwan's 1,187 billion pharmaceutical market, the domestic pharmaceutical market share only 22% in the Taiwan market small , market competition is very fierce, and the research and innovation is weak, R & D talent fault, nor is it a short time new drug development can be achieved, the U.S. FDA conservative in its attitude, making new drug approval has slowed down, the challenges of intellectual property rights, coupled with increasingly stringent regulatory review , raw material medicine continued to grow in recent years, the financial turmoil and the National Health Insurance drug prices lowered significantly the double blow, making revenue and profits hit the pharmaceutical industry's business threatened.
　　From the relevant literature review, pointing out that customer value is within a certain period of time the potential contribution of an enterprise customer, the case study company for the pharmaceutical industry, its customers including hospitals, clinics and pharmacies of medication officers belong to the professional and technical personnel , the patient will usually accept the recommendations of the professional doctors or pharmacists to the types of drugs, thus changing the pharmaceutical industry in a competitive environment If you want to come to the fore, with the physician or pharmacist to establish good customer relationships, create customer value, they are important.
The purpose of this study, expectations from the customer database to explore its trading characteristics of the customer grouping, and then made for the various groups of customers marketing proposals and appropriate management of meaning, expectation, managers will be better use of limited resources, and to assist management decision-making reference.
　　Based on the above, this study sample of domestic pharmaceutical companies to conduct customer value analysis, customer data and sales information, the use of customer value analysis (RFM) and the K-number method (K-mean method), to be grouping the customer, their Input variables for the RFM, the recent purchase (R), purchase frequency (F) and the average purchase amount (M) and total purchase amount (TM), and the results of two kinds of methods to compare results of the study RFM analysis method, the customer is divided into 13 groups, from the data found that all the grouping is not obvious and the similarity between the characteristics of a very high, and the other using Microsoft SQL Server 2005 Analysis Services (SSAS) analysis, the customer is divided into 10 groups, 10 groups are also using the system default values , you can avoid too many sub-groups, resulting in scattered resources, the study also found that K-mean model analysis of the effects of a better and more able to see the various groups of transactions between the different characteristics, finally, sum up in 10 different investment behavior of the cluster, do The name for the cluster, then the cluster clients for different features to develop a different marketing strategy, hoping to provide real-capital management decisions on the reference.

並列關鍵字

cluster analysis ； customer relationship management (CRM) ； United States Food and Drug Administration (FDA) ； customer value (RFM) ； data mining

參考文獻

[24]李淑芳，王鐵人，蔡源成(2006)，網路市場導向對網路顧客忠誠度之影響－以關係品質為中介，中華管理學報，1-16頁。

[29]陳建州(2006)，以群集分析輔助顧客終身價值之計算，碩士論文，元智大學資訊管理學系。

[6] 陳麗君(2003)，應用資料探勘技術於信用卡黃金級客戶之顧客關係管理。

[11] 黃靖如(2004)，-醫院病患服務之研究-以CRM觀點。

[14] 李連滋(1997)，「中華民國化學工業年鑑－製藥工業」，台灣經濟研究院。