臺灣博碩士論文加值系統

在通信業激烈的競爭下，為了維護公司營利，通信業者必須防止用戶流失，亦即因用戶轉換電信公司所減少的獲利。本論文提出ㄧ個智慧型客戶保留管理系統來處理客戶保留之工作。首先，本系統分析歷史用戶資料庫來學得用戶流失率的預測模組，並將此模組最佳化。本模組可用來預測任一輸入用戶之流失率。本系統亦建構ㄧ策略模組，將保留策略知識本體中適當的保留策略，對應到據特定的關聯屬性值群組。本模組可在判斷潛在流失客戶的價值後，藉由分析客戶的特性，自動提供適當的保留策略來吸引高附加價值的客戶。建構本模組所需之保留策略知識本體係廣泛收集與分析業界及學術界之相關客戶保留策略所建構而成，而關聯屬性値群組則是分析從歷史流失客戶資料庫所挖掘出來的屬性値間的關係而來。我們的實驗結果顯示，用戶流失率預測模組可以達到百分之八十五的準確度。此外，跟某ㄧ電信業者網佔所提供之費率方案估算系統比較，由本系統自動建議潛在流失客戶的費率方案與該系統所提供的費率方案也幾乎完全吻合。
本論文的貢獻主要分有三： 第一，本論文所提出的系統確可依照潛在流失客戶不同的特性，自動建議適當的保留策略。第二，我們所挖掘出來的流失客戶關聯屬性値群顯示了重要的資訊，確可幫助我們分析流失客戶的行為模式。最後，我們所建構的保留策略知識本體不僅提供了本系統所需的多元化保留策略，並且可以支援ㄧ般性保留策略的設計與分析。

Competition in the wireless telecommunications industry is fierce. To maintain profitability, carriers must control churn, meaning the loss of subscribers who defects from one carrier to another. This thesis proposes an intelligent customer retention management system to deal with the customer retention problem. The system first learns a churn prediction model by the decision tree-based technique from a historical subscriber database. The churn model is then optimized to support the prediction of defection probability of a customer. The system also constructs a policy model from mapping the analyzed churn attribute clusters to the retention policy ontology. The policy model supports automatic proposal of suitable retention policies to retain a potential churner provided that she is a valuable subscriber. Our experiment shows the learned churn model has around 85% of correctness in tenfold cross-validation. And a preliminary test on proposing suitable package plans shows the policy model works equally well as a commercial website.
The contributions of the work are as follows. First, the churn attribute clusters we discovered show some significant relationships among the attribute values, which can reveal why a customer defects more effectively. Second, the retention policy ontology we constructed not only can support the construction of the policy model in our system but also can support general retention policy design and analysis. Finally, the fact that the system can automatically propose proper retention policies for potential churners according to their specific characteristics is new and important in customer retention study.

摘要 i
Abstract ii
誌謝 iii
TABLE OF CONTENTS iv
LISTS OF FIGURES vi
LISTS OF TABLES vii
Chapter 1 Introduction 1
1.1 Motivation 1
1.2 Problem Specification 4
1.3 Contributions 4
1.4 Thesis Organization 5
Chapter 2 Related Work 6
2.1 Data Mining 6
2.1.1 Definition 6
2.2 Classification Analysis 8
2.3 Prediction Analysis 11
2.4 Decision Tree-based Algorithms 13
2.4.1 Information Gain Criterion 16
2.4.2 Gain Ratio Criterion and C4.5, C5.0 17
2.4.3 Cross-Validation 18
2.5 Association Rule Mining Techniques 19
2.5.1 Apriori Algorithm 20
2.6 Ontology 21
Chapter 3 System Architecture 23
3.1 System Architecture 23
3.2 Churn Prediction 25
3.2.1 Historical Subscriber Database 25
3.2.2 Churn Model Learner 28
3.2.3 Churn Model 30
3.2.4 Model Optimization 31
3.3 Policy Generation 32
3.3.1 Churn Attribute Cluster Constructor 32
3.3.2 Retention Policy Ontology 35
3.3.3 Policy Model Constructor 40
3.4 Application Mode 43
Chapter 4 System Implementation and Evaluation 45
4.1 Churn Model Implementation 45
4.1.1 Model Construction 45
4.1.2 Model Optimization 48
4.2 Policy Model Implementation 52
4.2.1 Churn Attribute Cluster Model Construction 52
4.2.2 Retention Policy Ontology Construction 53
4.2.3 Knowledge Mapping 54
4.3 System Operation 57
4.4 System Evaluation 58
Chapter 5 Conclusions and Future Work 60
5.1 Conclusions 60
5.2 Comparison with [Tsai02] 61
5.3 Contribution 62
5.4 Discussion and Future Work 63
References 65
作者簡介 67

[Agra94]R. Agrawal and R. Srikant, “Fast Algorithms for Mining Association Rules,” Proceedings of the 20th International Conference on Very Large Databases, pp. 487-499, 1994.
[Au03A]Wai-Ho Au and Keith CC Chan, “Mining Fuzzy Association Rules in a Bank-Account Database,” IEEE Transactions on Fuzzy Systems, Vol. 11, No. 2, pp. 238-248, APRIL 2003.
[Au03B]Wai-Ho Au, Keith CC Chan, and Xin Yao, Fellow, “A Novel Evolutionary Data Mining Algorithm With Applications to Churn Prediction,” IEEE Transactions on Evolutionary Computation, Vol. 7, No. 6, pp. 532-545, DECEMBER 2003.
[Berr00]Michael J. A. Berry and Gordon S. Linoff, “Mastering Data Mining: The Art and Science of Customer Relationship Management”, John Wiley & Sons, Inc., 2000.
[Dehn01]F. Dehne, T. Eavis and A. Rau-Chaplin, “Computing Partial Data Cubes for Parallel Data Warehousing Applications,” Proceedings of Euro PVM/MPI 01, Santorini, Greece, 2001.
[Dunh03]M. H. Dunham, “Data Mining Introduction and Advanced Topics”, Prentice Hall, 2003.
[Fens01]D. Fensel, Ontologies: A Silver Bullet for Knowledge Management and Electronic Commerce, Springer, 2001.
[Gerp01]Gerpott, TJ, W. Rambs and A. Schindler, “Customer retention, loyalty, and satisfaction in the German mobile cellular telecommunications market,” Telecommunications Policy, Volume 25, pp. 249-269, 2001.
[KDnu05]KDnuggets Statistical Report. Available at: http://www.KDnuggets.com.
[Kim04] Hee-Su KIM and Choong-Han YOON, “Determinants of subscriber churn and customer loyalty in the Korean mobile telephony mrket,” Telecommunications Policy, Volume 28, Issue 9-10, Pages 751-765, 2004.
[Moze00]M. C. Mozer, R. Wolniewicz, and D. B. Grimes, “Predicting Subscriber sDissatisfaction and Improving Retention in the Wireless Telecommunications Industry,” IEEE Transactions on Neural Networks, Vol. 11, No. 3, pp. 690-696, MAY 2000.
[Oceans]Oceansblue, “Accurate Customer Churn Prediction with Information Inventor”. Available at: http://www.oceansblue.co.uk/resources/iichurn1.pdf.
[Quin83]J. R. Quinlan, “Induction of Decision Tree,” Machine Learning, Vol.1, pp. 81-106, 1983.
[Quin93] J. R. Quinlan, “C4.5: Programs for Machine Learning,” Morgan Kaufmann, San Mateo CA, 1993.
[Ross03]Saharon Rosset and Einat Neumann, “Integrating Customer Value Considerations into Predictive Modeling”, Proceeding of the Third International Conference on Data Mining, pp. 283-290, 2003.
[SPSS99]SPSS Inc. (1999). “Working with Telecommunications: Churning In the Telecommunications Industry”. SPSS White Paper, available at: http://www.spss.com/downloads/papers/.
[SPSS04]SPSS Inc. (2004). “Five Predictive Imperatives for Maximizing Customer Value”. SPSS White Paper, available at: http://www.spss.com/downloads/papers/.
[Tsai02]M.S. Tsai, B.H.Chu, and C.S. Ho, “A Hybrid Data Mining Architecture for Customer Retention,” Proc. International Computer Symposium (ICS-2002), Hua-Lian, Taiwan, pp. 1838-1844, 2002.12.18.
[Usch96]M. Uschold and M. Gruninger, “ONTOLOGIES: Principles, Methods and Applications,” The Knowledge Engineering Review, vol. 11, no. 2, pp. 93-136, Feb. 1996.
[Wei02]Wei, C. and Chiu, I., “Turning Telecommunications Call Details to Churn Prediction: A Data Mining Approach,” Expert Systems with Applications, Vol.23, No. 2, 2002, pp.103-112.
[Yan04]L. Yan, Richard H. Wolniewicz, Robert Dodier. "Predicting Customer Behavior in Telecommunications," IEEE Intelligent Systems, vol. 19, no. 2, pp. 50-58, March/April 2004.
[Yan01]L. Yan , et al., "Improving Prediction of Customer Behavior in Nonstationary Environments," Proc. Int'l Joint Conf. Neural Networks, IEEE/Omnipress, vol.3, 2001,pp. 2258–2263.
[Yin03]Xiaoxin Yin, Jiawei Han. “CPAR: Classification based on Predictive Association Rules,” 3rd SIAMInternational Conference on Data Mining (SDM'03), San Francisco, CA, June, 2003.
[Zhen01]Z. Zheng, R. Kohavi and L. Mason, “Real world performance of association rule algorithm,” Proc. 7-th Int. Conf. on Knowledge Discovery and Data Mining, 2001.