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。本論文提出一個混合式的架構來完成客戶保留之工作，此架構非但能預測用戶流失的可能性，並且提出保留策略。
本架構有兩種執行模式，即學習模式與應用模式。在學習模式中，流失模式學習器從用戶歷史紀錄資料庫中分析並學得潛在的關聯，以建立流失模式模組。接下來策略模組建構器便依照出現於流失模式模組中的屬性，將所有的流失客戶分成不同的群組，同時也針對每組流失用戶群建構合適的特定策略模組。在應用模式中，流失用戶預測器使用流失模式模組預測某用戶流失的可能性，若其可能性偏高，便引入策略制定器依據策略模組來提供特定保留策略。從我們的實驗結果顯示，已建構完成的流失模式學習器大約有百分之八十五的正確性，雖目前無適當資料來評估已建構的策略模組，然而從建構的過程中，我們發現一些有趣而重要的做法，能夠對保留可能流失的用戶提供較好的幫助。
本研究之所以重要，是因現有的研究都只致力於如何提高流失預測的精確度，從未討論到保留策略或只是根據決策樹的路徑建議保留策略。我們的策略模組建構器則深入探討流失用戶群的觀念，相當於發掘決策路徑間的關聯。我們相信對於流失用戶間的關聯有更深入的了解，更能依此提出較佳的保留策略模組。

Competition in the wireless telecommunications industry is fierce. To maintain profitability, wireless carriers must control churn, which is the loss of subscribers who switch from one carrier to another. This thesis proposes a hybrid architecture that tackles the complete customer retention problem, in the sense that it not only predicts churn probability but also proposes retention policies. The architecture works in two modes, namely, the learning and usage modes. In the learning mode, the churn model learner learns potential associations inside the historical subscriber database to form a churn model. The policy model constructor then uses the attributes that appear in the churn model to segment all churners into distinct groups. It is also responsible for developing a specific policy model for each churner group. In the usage mode, the churner predictor uses the churn model to predict the churn probability of a given subscriber. A high churn probability will cause the churner predictor to invoke the policy maker to suggest specific retention policies according to the policy model. Our experiments illustrate that the learned churner model has around 85% of correctness in evaluation. Currently, we have no proper data to evaluate the constructed policy model. The construction process, however, signifies an interesting and important approach toward a better support in retaining possible churners.
This work is significant since the state-of-the-art technology only focuses on how to increase the accuracy of churn prediction. They either never touched the issue of retention policies, or only proposed policies according to the path conditions of the decision tree, the churn model. Our policy model construction process goes on step further to investigate the concept of churner groups, which equivalently digs out the associations between the paths of the decision tree. We believe with this in depth knowledge about how churns are related, we can propose better retention policy models for possible churners.

ABSTRACT (IN CHINESE) i
ABSTRACT (IN ENGLISH) iii
ACKNOWLEDGEMENT (IN CHINESE) v
TABLE OF CONTENTS vi
LIST OF TABLES viii
LIST OF FIGURES ix
CHAPTER 1 INTRODUCTION 1
1.1 MOTIVATION 1
1.2 PROBLEM SPECIFICATION AND OUR APPROACH 4
1.3 RELATED WORK 5
1.4 ORGANIZATION OF THE THESIS 6
CHAPTER 2 RELATED TECHNOLOGY 8
2.1 DATA MINING 8
2.2 CLASSIFICATION AND PREDICTION 10
2.3 CLUSTERING ANALYSIS 19
TABLE OF CONTENTS
ABSTRACT (IN CHINESE) i
ABSTRACT (IN ENGLISH) iii
ACKNOWLEDGEMENT (IN CHINESE) v
TABLE OF CONTENTS vi
LIST OF TABLES viii
LIST OF FIGURES ix
CHAPTER 1 INTRODUCTION 1
1.1 MOTIVATION 1
1.2 PROBLEM SPECIFICATION AND OUR APPROACH 4
1.3 RELATED WORK 5
1.4 ORGANIZATION OF THE THESIS 6
CHAPTER 2 RELATED TECHNOLOGY 8
2.1 DATA MINING 8
2.2 CLASSIFICATION AND PREDICTION 10
2.3 CLUSTERING ANALYSIS 19
CHAPTER 3 HYBRID CUSTOMER RETENTION SYSTEM 27
3.1 SYSTEM ARCHITECTURE 27
3.2 CHURN MODEL LEARNER 28
3.3 CHURN MODEL 31
3.4 POLICY MODEL CONSTRUCTOR 32
3.5 USAGE MODE 35
CHAPTER 4 EXPERIMENTAL RESULTS 38
4.1 TRAINING DATABASE AND CHURN MODEL 38
4.2 CHURN GROUPS AND POLICY MODEL 42
4.3 DISCUSSION 47
CHAPTER 5 CONCLUSIONS AND FUTURE WORK 49
5.1 CONCLUSIONS 49
5.2 FUTURE WORK 50
REFERENCES 53
LIST OF TABLES
Table 2.1 Difference between decision tree algorithm 13
Table 2.2 Categories of clustering algorithm 21
Table 4.1 Results of different combinations of database partitions 39
Table 4.2 Attributes appearing in churn model 42
Table 4.3 Policy model (only specific policies shown) 45
LIST OF FIGURES
Figure 1.1 What Are the Primary Reasons Customers Defect? 4
Figure 2.1 Example decision tree 14
Figure 2.2 Insertion of units to a SOM 25
Figure 2.3 Example of a trained GHSOM 26
Figure 3.1 System architecture in learning mode 27
Figure 3.2 System architecture in usage mode 28
Figure 3.3 File defining subscriber classes and attributes 31
Figure 3.4 A snapshot of part of churn model 32
Figure 3.5 Example of a GHSOM segmented cluster 33
Figure 3.6 Customer Segmentation Matrix for Wireless Carrier 34
Figure 3.7 Policy model for the churner group in Fig. 3.5 35
Figure 3.8 Enter subscriber data to the churner predictor 36
Figure 3.9 Churn probability predicted by the churner predictor 36
Figure 3.10 Proposed policies by the policy maker 37
Figure 4.1 Churn model in terms of decision tree 40
Figure 4.2 Preliminary churn groups by GHSOM using attributes in the churn model………………………………………………………………..……..…43
Figure 4.3 Churner groups 44
Figure 4.4 Preliminary churn groups by GHSOM using all attributes 48
Figure 4.5 Churner groups on all attributes 48

[Ande2000] Anderson Consulting, “Battling Churn to Increase Shareholder Value: Wireless Challenge for the Future”, Anderson Consulting Research Report, 2000.
[Bers1999] S. Berson, K. Thearling, and S. J. Smith, Building Data Mining Applications for CRM. McGraw-Hill Professional Publishing, 1999.
[Boun2001] C. Bounsaythip and E. R. Runsala, “Overview of Data Mining for Customer Behavior Modeling”, VTT Information Technology research report, June 2001.
[Cest1987] B. Cestnik, I. Kononenko, and I. Bratko, “ASSISTANT 86: A Knowledge-Elicitation Tool for Sophisticated Users”, Proceedings of Progress in Machine Learning, pp. 31-45, 1987.
[Chan2000] W. L. Chang, A Synthesized Learning Approach for Web-Based CRM, Master Thesis, Fu Jen University, Taipei Taiwan, 2000.
[Chiu2000] I-T. Chiu, Telecommunications Data Mining for Churn Prediction, Master Thesis, National Sun Yat-sen University, Kaohsiung Taiwan, 2000.
[Dill1995] R. Dilly, “Data Mining - An Introduction”, Available at http://www.pcc.qub.ac.uk/tec/courses/datamining/stu_notes/dm_book_1.html, December 1995.
[Ditt2000a] M. Dittenbach and A. Rauber, “The Growing Self-Oranizing Map (GHSOM) - Introduction and Architecture”, Available at http://www.ifs.tuwien.ac.at/~mbach/ghsom/, 2000.
[Ditt2000b] M. Dittenbach, D. Merkl, and A. Rauber, “The Growing Hierarchical Self-Organizing Map”, Proceedings of the International Joint Conference on Neural Networks (IJCNN 2000), pp. 15-19, July 2000.
[Gerp2001] T. J. Gerpott, W. Rams, and A. Schindler, “Customer Retention, Loyalty, and Satisfaction in the German Mobile Cellular Telecommunications Market”, Telecommunications Policy, Vol. 25, pp. 249-269, 2001.
[Howl2000] D. Howlett, “That Crazy Little Thing Called Churn”, Boardwatch Magazine, Vol. 3, pp. 23-45, April 2000.
[Kamb2000] M. Kamber and J. Han, Data Mining: Concepts and Techniques, Morgan Kaufmann Publishers, August 2000.
[Koho1982] T. Kohonen, “Self-organized Formation of Topologically Correct Feature Maps”, Biological Cybernetics, pp. 59-69, 1982.
[Yan2001] L. Yan, D. J. Miller, M. C. Mozer, and R. Wolniewicz, “Improving Prediction of Customer Behavior in Nonstationary Environments”, Proceedings of International Joint Conference on Neural Network, Vol. 3, pp. 2258-2263, 2001.
[INDI2002] INDIGOlighthouse company, “Customer Relationship Management”, Available at http://www.indigolighthouse.com/crm.htm, 2002.
[Merk2000] D. Merkl, M. Dittenbach, and A. Rauber, “Using Growing Hierarchical Self-Organizing Maps for Document Classification”, Proceedings of BSANN Proceeding-Europan Syposium on Artificial Neural Networks, Bruges, pp. 7-12, April 2000.
[Modi1999] L. Modisette, “Milking Wireless Churn for Profit”, Telecommunications Online, Web page: http://www.telecoms-mag.com/default.asp, February 1999.
[Moze2000] M. C. Mozer, R. Wolniewicz, and D. B. Grimes, “Predicting Subscriber Dissatisfaction and Improving Retention in the Wireless Telecommunications Industry”, IEEE Transactions on Neural Networks, Vol. 11, No. 3, pp. 690-696, MAY 2000.
[Ng1999] K. S. Ng and H. Liu, Customer Retention via Data Mining, Kluwer Academic Publishers, 1999.
[Quin1983] J. R. Quinlan, “Induction of Decision Tree”, Machine Learning, Vol.1, pp. 81-106, 1983.
[Quin1993] J. R. Quinlan, C4.5: Programs for Machine Learning, Morgan Kaufmann, San Mateo CA, 1993.
[Raub2001] A. Rauber, M. Dittenbach, and D. Merkl, “Towards Automatic Content-Based Organization of Multilingual Digital Libraries”, Proceedings of the 3rd All-Russian Scientific Conference, Proceedings of Digital Libraries: Advanced Methods and Technologies, Digital Collections (RCDL 2001), September 2001.
[Schm1999] J. Schmitt, “Churn: Can Carriers Cope?”, Telecommunication Online, Web page: http://www.telecoms-mag.com/default.asp, February 1999.
[Srin2000] M. Srinivasan, “Keeping Your Customers”, Available at http://research.badm.sc.edu/research.edu/, December 2000.
[SAS2001] SAS Company, “Predicting Churn: Analytical Strategies for Retaining Profitable Customers in the Telecommunications Industry”, A SAS White Paper, 2001.
[Schw2001] E. Schweighofer, A. Rauber, and M. Dittenbach, “Automatic Text Representation, Classification and Labeling in European Law”, Proceedings of the 8th International Conference on Artificial Intelligence and Law (ICAIL 2001), pp. 21-25, May 2001.