電信數據電路客戶流失資料探勘

Churn of Telecom Data Circuit Using Data Mining Technique

葉燉烟(Duen-Yian Yeh)；鄭景俗(Ching-Hsue Cheng)；傅雲龍(Yun-Long Fu)

顧客流失 ； 資料探勘 ； 叢集分析 ； 類神經叢集化 ； Churn ； Data Mining ； Clustering Analysis ； Neural Clustering

管理與系統

12卷2期（2005 / 04 / 01）

75 - 91

繁體中文

民國九十年二月起交通部陸續核發三家（台灣固網、速博電信、東森寬頻電信）固定通信網路經營特許執照，加上原有之中華電信共計有四家均可經營第一類電信業務，包括出租各種速率之數據電路供應國內市場需求。因而整個電信市場瞬間形成多家激烈競爭的局面，業者紛紛推出各種優惠促銷方案來拉攏客戶。各家業者為維持市場佔有率及降低日趨嚴重的客戶流失問題，均不惜降低利潤來保有客戶。據統計，企業得到一位新客戶所花的成本是維繫一位老客戶的五至十倍代價。因此業者莫不想盡辦法來留住老客戶，同時也對客戶的消費行為進行探勘分析，企圖找出可能流失的客戶並及早速謀新的行銷策略。本研究主要是應用叢集分析(Cluster Analysis)技術，以IBM Intelligent Miner及MATLAB為資料探勘(Data Mining)工具，分別將國內某電信業者的數據電路退租資料進行「類神經叢集化」(Neural Clustering)、「人口統計叢集化」(Demographic Clustering)、「模糊C均值叢集化」(Fuzzy C-Means Clustering)探勘分析，進而選取較佳之叢集分析結果做合理推論，以探勘出可能流失的客戶及哪些產品在競爭市場上已漸居劣勢，必須速謀因應對策。經由探勘分析，本文發現：⑴Group2的16565位客戶為獲利性較低的高忠誠客戶。⑵Group3的31位客戶是具有高度流失傾向的大企業。⑶Group4的319位客戶為高收益且移轉性低的高忠誠中小企業。在本研究預測出的31個高潛在流失客戶中，經過印證實際潛在流失客戶數後，發現有23個客戶確實已與其他同業進行新合約的協商，印證準確率達74%。這些經資料探勘的成果，將對電信固網業者之數據電路經營具有提昇競爭力的加分效果。

The Ministry of Transportation and Communications had issued three concession licenses of Type Ⅰ telecommunications to Eastern Broadband Telecom, Sparq Telecom, and Taiwan Fixed Network Telecom respectively from February 2001. Counting the original Chung-Hwa Telecom, there are four telecom companies permitted to operate Type Ⅰ telecommunications business, including renting out any rate of data circuit to meet the needs of the domestic market. As a result, fierce competition soon arises in the telecommunication market. In order to attract the clients, the providers come up with various favorable sale promotions, which make chaos in the telecommunication data circuit market. On the purpose of maintaining market share and reducing the downward problem of customer moving out, the providers choose to cut down profits to retain customers. According to statistics, however, the cost for a company to gain a new customer is as much as five to tell times more value than that of retaining an old customer. Therefore, the service providers spare no effort to retain old customers. Meanwhile, they try to analyze the attitudes of the customers with data mining technique, attempting to find out customers of the possible moving out and thus help companies to propose new marketing strategies. This study mainly applied the technique of Cluster Analysis, used IBM Intelligent Miller and MALAB as tools, and analyzed the customer's withdrawal data with Neural Clustering, Demographic Clustering, and Fuzzy C-Means Clustering, then selected a best one of chum data to infer the result reasonably. The purpose is to find out the possible moving out of customers and products that are going downward in the market, and thus launches strategies against the problems. These will give an effective help to the company decision maker, particularly obvious benefits and improvements on the marketing strategy planning and the performance. According to the results of data milling, this paper finds: ⑴ In Group2, 16565 customers are high loyal customers with low profitable. ⑵ There are 31 large enterprises with high shift tendency in Group3. ⑶ In Group4, 319 customers are high loyal firms with high profitable and low shift. These results will provide the conduct of data circuits of Telecom with the bonus effect of competitiveness improvement.

1. 張寶誠、楊景棠、袁建中、陳俊傑(2000)。台灣通訊工業發展趨勢預測與研究。管理與系統，7(2)，227-248。
   連結：
2. 張寶誠、楊景棠、袁建中、陳俊傑(2000)。台灣通訊工業發展趨勢預測與研究。管理與系統，7(2)，227-248。
   連結：
3. 簡禎富、李培瑞、彭誠湧(2003)。半導體製造資料特徵萃取與資料挖礦之研究。資訊管理學報，10(1)，63-48。
   連結：
4. Agrawal, R.,Lin, K.,Sawhney, H.S.,Shim, K.(1995).Fast Similarity Search in the Presence of Noise, Scaling and Translation in Time-Series Data bases.Proceedings of the 21st International Conference on Very Large Data Bases
5. Agrawal, R.,Srikant, R.(1995).Research Report RJ 9994, IBM Alma den Research Center.San Jose, California.
6. Agrawal, R.,Srikant, R.(1994).Fast Algorithms for Mining Association Rules.Proceedings of International Conference on Very Large Data Bases,487-499.
7. Berry, J.A.,Linoff, G.(1997).Data Mining for techniques: For Marketing Sale and Customer Support.Canada:John Wiley & Sons Inc..
8. Bezdek, J.C.(1981).Pattern Recognition with Fuzzy Objective Function Algorithms.New York.
9. Bohlen, M.,Bukauskas, L.,Eriksen, P.S.,Lauritzen, S.L.,Mazeika, A.,Musaeus, P.,Mylov, P.(2003).3D visual data mining- goals and experiences.Computational Statistics & Data Analysis,43,445-469.
10. Bolton, R.J.,Hand, D.J.,Crowder, M.(2004).Significance tests for unsupervised pattern discovery in large continuous multivariate data sets.Computational Statistics & Data Analysis,46,57-79.
11. Bult, J.R.,Wansbeek, T.(1995).Optimal selection for direct mail.Marketing Science,14(4),378-381.
12. Estivill-Castro, V.,Murray, A.T.(1997).Spatial Clustering for Data Mining with Generic Algorithms.Technical Report FIT-TR-97-10, Faculty of information Management
13. Fayyad, U.,Piatetsky-Shapiro, G.,Smyth, P.(1996).The KDD process for extracting useful knowledge from volumes of data.Communications of the ACM,39,27-34.
14. Han, J.,Kamber, M.(2000).Data Mining: Concepts and Techniques.Morgan Kaufmann.
15. Hand, D.J.(1998).Strategy, methods, and solving the right problem.Computational Statistics,13,5-14.
16. Schroedl, S.,Wagstaff, K.,Rogers, S.,Langley, P.,Wilson, C.(2004).Mining GPS traces for map refinement.Data Mining & Knowledge Discovery,9,59-87.
17. Schuster, A.,Wolff, R.(2004).Communication-efficient distributed mining of association rules.Data Mining & Knowledge Discovery,8,171-1996.
18. Sung, H.H.,Sang, C.P.(1998).Application of data mining tools to hotel data mart on the intranet for database marketing,1-31.
19. Zadeh, L.A.(1965).Fuzzy sets.Information and Control,8,338-353.
20. 毛治國(2002)。中華電信，36(1)
21. 林德國(2001)。口碑行銷。台北:遠流出版社。
22. 邱志洲、李天行、周宇超、呂奇傑(2002)。整合鑑別分析與類神經網路在資料探勘上之應用。工業工程學刊，19(2)，9-22。
23. 張德民(1999)。資料探勘：從搜尋金星火山到偵查考試作弊。資訊傳真周刊
24. 謝邦昌(2001)。資料採礦入門及運用。台北:資商訊息顧問股份有限公司。
25. 簡禎富、林鼎浩、劉巧雯、彭誠湧、徐紹鐘、黃佳琪(2002)。建構晶圓圖分類之資料挖礦。工業工程學刊，19(2)，23-38。