臺灣博碩士論文加值系統

近年來在網際網路和電子商務日益普及之下，網路購物已成為未來的趨勢，在企業經營方面，往往以利潤最大化為第一目標，要達到利潤最大化，公司就必須實施顧客關係管理。然而本論文是以網路購物顧客的屬性作為輸入單元，在潛藏狄立克雷分配模型裡利用變分近似推論預估參數來將顧客分群，找出高忠誠度、潛在性、無獲利性顧客，並且探討顧客在購買商品時，其購買次數和購買金額、對網路商店的感受和平常使用網路方面的程度。

In recent years, internet and electronic commerce become popular as consumers getting tendency toward online shopping. The new economic has been developed into the internet virtual space. A new economic era is born. Enterprises often set objectives on acquiring profit-maximizing in their businesses. To achieve at profit-maximizing, enterprises must enhance customer relationship management and adjust marketing strategy. It is necessary for enterprises to engage in process of customer relationship management analysis on their daily business. For online shopping market, useful information can be extracted from customer database which is beneficial from shopping website.

Our approach involves customer segmentation on customers’attributes which stored in multiple relations. By using latent Dirichlet allocation to segment customers, we can distinguish high loyal customers, potential customers, or losing customers. From customer segmentation, we explore customer value for e-company and online shopping channels. This information offers decision makers to improve their customer relationship management and adjust marketing strategy.

We employ latent Dirichlet allocation model into our approach. In LDA model, we present approximate inference techniques based on variational methods and an EM algorithm for empirical Bayes parameter estimation. Variational approximate provides an effective computing method to segment customers into segmentation.

According to customer-centric business tendency, we make suggestions for handling customers’ direction of purchase and pandering to consumer behavior or consumer needs. This allows an enterprise to achieve its objective by building up profitable relationships with customers and enhancing customer satisfaction and loyalty.

TABLE OF CONTENT
ABSTRACT i
TABLE OF CONTENT iii
LIST OF FIGURE vi
LIST OF TABLE vii
Chapter 1 Introduction 1
1.1 Background 1
1.2 Motivation 2
1.3 Objectives 3
1.4 Research Process 4
Chapter 2 Related Works 5
2.1 Customer Relationship Management 5
2.1.1 Web-based CRM 6
2.1.2 A Strategic Framework for CRM 7
2.2 Data Mining 7
2.3 Online Shopping 10
2.4 RFM 11
2.4.1 Stone’s RFM Model 12
2.4.2 Arthur Hughes’ RFM Model 13
2.4.3 Customer Quintiles 13
2.4.4 Behavior Quintile Scoring 15
2.4.5 Conclusion of RFM Model 16
Chapter 3 Spectral Relational Clustering and Latent Class Representation of GoM Model 17
3.1 Graph Partitioning 17
3.1.1 Spectral Clustering Algorithm 18
3.1.2 Weighted Kernel K-means 19
3.2 Spectral Relational Clustering 20
3.2.1 K-partite Graphs for Relational Data 21
3.2.2 Model Formulation 21
3.2.3 Spectral Relational Clustering 23
3.3 Latent Class Representation of the Grade of Membership Model 24
3.3.1 Latent Class Model 25
3.3.2 Grade of Membership Model 26
Chapter 4 Latent Dirichlet Allocation 30
4.1 Dirichlet Distribution 30
4.2 Latent Dirichlet Allocation 32
4.2.1 Procedure for Inference and Parameter Estimation under LDA 33
4.2.2 Variational Inference 34
Chapter 5 Research Methodology 36
5.1 Multiple Relational Data Sets 37
5.2 Graphical Model Representation of Latent Dirichlet Allocation 38
Chapter 6 Variational Approximation 45
6.1 Graphical Model of Variational Approximation 45
6.2 Expectation-Maximization Algorithm 48
Chapter 7 Experiments Design and Experimental Results 51
7.1 The experiment 51
7.2 The experiment 53
7.2.1 Soft Clustering Approach 54
7.2.2 Characteristic of Customers in Each Segment 60
7.3 The experiment 62
7.3.1 Summary of Segments on Attributes of Demography Data 63
7.3.2 Summary of Segments on All Attributes of Demography Data 65
7.4 The experiment 66
7.5 The experiment 69
7.6 Marketing Strategy in Each Segment 70
7.7 The experiment 72
Chapter 8 Conclusions 74
References 76
Appendices 81
Appendix A 網路購物之消費行為研究過錄編碼簿 81
Appendix B.1 A Codebook of Consumer Behavior for Online Shopping 95
Appendix B.2 Multiple Relations Data Sets and Their Attributes 99
Appendix C.1 Hard Clustering (K=5) 100
Appendix C.2 Soft Clustering (K=5) 100
Appendix D Online Service Response and Internet Usage of Segments 102
Appendix E Percentage of Customers for Each Segment in Demography 103
Appendix F Correlative in Terns of Customers between Segments and Weka Clusters 105




LIST OF FIGURE
Figure 1.1 Research Process 4
Figure 2.1 A Conceptual Framework for CRM Strategy 8
Figure 2.2 Equal Numbers of Customers in Each Group 14
Figure 2.3 The Mean Methods Yield Sensitivity at Both Top and Bottom but also Isolates Single Purchasers 15
Figure 4.1 Graphical Model of Latent Dirichlet Allocation 34
Figure 4.2 Variational Approximation Model for Latent Dirichlet Allocation 35
Figure 5.1 Research Framework 36
Figure 5.2 Multiple Relational Data Sets 37
Figure 5.3 Graphical Model Representation of Latent Dirichlet Allocation 39
Figure 6.1 Graphical Model Representation of Variational Approximation 47
Figure 7.1 Expectation Values in Each Iteration for Hard Clustering with K=5 52
Figure 7.2 Expectation Values in Each Iteration for Soft Clustering with K=5 53
Figure 7.3 Customer Distribution in Each Segment 57
Figure 7.4 Customer Counts of Buying Frequency in Each Segment 58
Figure 7.5 Customer Counts of Monetary Spending in Each Segment 58
Figure 7.6 Pie Chart of Segments Statistic Information 59




LIST OF TABLE
Table 7.1 Relationship Matrices between Clusters of Relations 54
Table 7.2 Results of Segmentation 55
Table 7.3 Customers per Segment of Hard Clustering v.s Soft Clustering (K = 5) 56
Table 7.4 Summary of Nine Segments 57
Table 7.5 Age Distribution for Each Segment 63
Table 7.6 Gender Distribution for Each Segment 64
Table 7.7 Income Distribution for Each Segment 64
Table 7.8 Marriage Distribution for Each Segment 65
Table 7.9 Major Customers for Each Segment of Demography Data 66
Table 7.10 Significant Customers for Each Segment of Demography Data 66
Table 7.11 Hard Clustering v.s Soft Clustering (K = 5) 67
Table 7.12 Hard Clustering v.s Soft Clustering (K = 10) 67
Table 7.13 Hard Clustering v.s Soft Clustering (K = 20) 68
Table 7.14 Hard Clustering v.s Soft Clustering (K = 30) 68
Table 7.15 Number of Iterations 69
Table 7.16 5% Difference v.s 10% Difference v.s 15% Difference 70
Table 7.17 Number of Customers for Each Cluster 72
Table 7.18 Significant Customers for Each Cluster of Demography Data 73

[1] E. M. Airodi, D. M. Blei, S. E. Fienberg and E. P. Xing, “Mixed Membership Stochastic Block Models for Relational Data with Application to Protein-Protein Interactions”, Biometrics Society Annual Meeting, 2006.
[2] Hagai Attias, “A Variational Baysian Framework for Graphical Models”, In Advances in Neural Information Processing Systems 12, 2000.
[3] Francis R. Bach and Michael I. Jordan, “Learning spectral Clustering”, In Proc. of NIPS-16. MIT Press, 2004.
[4] K. Barnard, P. Duygulu, de Fraitas, N. Forsyth, D. Blei, and Michael I. Jordan, “Matching words and pictures”, Journal of Machine Learning Research, 3:993-1022, 2003.
[5] D. M. Blei, A. Ng and Michael I. Jordan, “Latent Dirichlet Allocation”, Journal of Machine Learning Research, 3:993-1022, 2003.
[6] D. M. Blei, Thomas L. Griffiths, Michael I. Jordan and Joshua B. Tenenbaum, “Hierarchical Topic Models and the Nested Chinese Restaurant Process”, Advances in Neural Information Processing Systems, 2004.
[7] David M. Blei, Michael I. Jordan and Andrew Y. Ng, “Hierarchical Bayesian Models for Applications in Information Retrieval”, Bayesian Statistics, 7:pp. 25-43, 2003.
[8] J.R. Bult, and T.J. Wansbeek, “Optimal selection for direct mail”, Marketing Science 14, 1995, pp. 378-394.
[9] I.S. Dhillon, Y. Guan, and B. Kulis, “Kernel k-means, Spectral Clustering and Normalized Cuts”, Proc. 10th ACM SIGKDD Int''l Conf. Knowledge Discovery and Data Mining, 2004.
[10] I. Dhillon, Y. Guan, and B. Kulis, ”A Unified View of Kernel k-means, Spectral Clustering and Graph Cuts”, Technical Report TR-04-25, University of Texas at Austin, 2005.
[11] Chris Ding, XiaoFeng He, Hongyuan Zha, Ming Go, and Horst Simon, “A Min-Max Cut Algorithm for Graph Partitioning and Data Clustering”, Proc. of International Conference on Data Mining, San Jose, CA, USA , 2001, pp 107-114.
[12] Elena A. Erosheva, “Partial Membership Models with Application to Disability Survey Data”, In H. Bozdogan (Ed.): New Frontiers of Statistical Data Mining, Knowledge Discovery, and E-Business, CRC Press, Boca Raton, 117-134, 2003b.
[13] Elena A. Erosheva, “Bayesian Mixed Membership Models for Soft Classification”, The Ubiquitous Challenge, pages 11--26. Springer-Verlag, 2005.
[14] Elena A. Erosheva, “Bayesian Estimation of the Grade of Membership Models”, Bayesian Statistics, In Bayesian Statistics, 7 . UK: Oxford Univ. Press, 2003.
[15] Jonathan Huang, “Maximum Likelihood Estimation of Dirichlet Distribution parameters”, Manuscript, 2006.
[16] A.M. Hughes, “Boosting response with RFM”, Marketing Tools, Vol.5, 1996, pp. 4-10.
[17] T. Jiang, and A. Tuzhilin, “Segmenting Customers from Population to Individual: Does 1-to-1 Keep Your Customers Forever?” IEEE TKDE, Vol. 18, No. 10, October, 2006.
[18] SD. Kamvar, D. Klein, and CD. Manning, “Spectral learning”, In Proc. 17th International Joint Conference on Artificial Intelligence (IJCAI), 2003.
[19] Charles Kemp, Thomas L. Griffiths and Joshua B. Tenenbaum, “Discovering Latent Classes in Relational Data”, Technical Report AI Memo 2004--019, MIT, 2004.
[20] D. D. Lee, and H. S. Seung, “Learning the parts of objects by non-negative matrix
factorization”, Nature, Vol. 401, 1999, pp. 788-791.
[21] B. Long, Z. Zhang, and P. S. Yu, “Co-clustering by block value decomposition”, in proc. ACM SIGKDD, 2005.
[22] Bo Long, Xiaoyun Wu, Zhongfei (Mark) Zhang, and Philip S. Yu, “Unsupervised learning on k-partite graphs”, SIGKDD, 2006.
[23] Bo Long, Zhongfei (Mark) Zhang, Xiaoyun Wu, and Philip S. Yu, "Spectral Clustering for Multi-Type Relational Data", Proc. International Conference on Machine Learning, ACM Press, Pittsburgh, PA, USA, June, 2006.
[24] Bo Long, Zhongfei (Mark) Zhang, and Philip S. Yu, “A Probabilistic Framework for Relational Clustering”, Proc. the 13th ACM International Conference on Knowl26edge Discovery and Data Mining, San Jose, CA, USA, August, 2007.
[25] Ulrike von Luxburg, “A Tutorial on Spectral Clustering”, Max Planck Institute for Biological Cybernetics, Technical Report No, TR-149, 2006.
[26] John A. McCarty, and Manoj Hastak, “Segmentation approaches in data-mining: A comparison of RFM, CHAID, and logistic regression”, Journal of Business Research 60, 2007, pp. 656-662.
[27] Jonh R. Miglautsch, “Thoughts on RFM Scoring”, Journal of Databases Marketing, Vol. 8(1), 2000, pp. 67-72.
[28] John R. Migkautsch, “Application of RFM principles: what to do with 1-1-1 customers”, Journal of Database Marketing, Vol. 9, No. 4, September, 2002, pp. 319-324.
[29] Thomas P. Minka, “Using lower bounds to approximate integrals”, Tech. Rep, Carnegie Mellon University, 2001.
[30] Thomas P. Minka, “Expectation-Maximization as lower bound maximization”, Technical report, MIT, 1998.
[31] A. Y. Ng, M. I. Jordan, and Y. Weiss, “On spectral clustering: Analysis and an algorithm”, In Proceedings of 14th Advances in Neural Information Processing Systems, 2001.
[32] Adrian Payne, and Pannie Frow, “A Strategic Framework for Customer Relationship Management”, Journal of Marketing, Vol. 69, October, 2005, pp. 167-176.
[33] Chris Regielski, Jyun-Cheng Wang, and David C. Yen, “Data Mining Techniques for Customer Relationship Management”, Technology in Society, Vol. 24, No. 4, November, 2002, pp. 483-502.
[34] B. Stone, “Successful Direct Marketing Methods”, Lincolnwood, NTC Business Books, IL, 1995.
[35] Bob Stone, and Ron Jacobs, “Successful direct marketing methods”, 2001, seventh edition, NTC Business Books.
[36] Bo Wang and D. M. Titterington, “Local convergence of variational Bayes estimators for mixing coefficients”, Technical Report 03-4, University of Glasgow, 2003.
[37] Jim Wheaton, “RFM cells: The ‘‘Kudzu’’ of segmentation”, DM News, 15th July, 1996.
[38] Jim Wheaton, “Superiority of tree analysis over RFM: How it enhances regression”, DM News, 12th August, 1996, pp. 21-23.
[39] Amoy X. Yang, “How to develop new approaches to RFM segmentation”, J Target Meas Anal Mark 13, 2004, pp. 50-60.
[40] Alex R. Zabah, Danny N. Bellenger, and Wesley J. Johnston, “An evaluation of divergent perspectives on customer relationship management: Towards a common understanding of an emerging phenomenon”, Industrial Marketing Management 33, 2004, pp. 475-489.
[41] Survey Research Data Archive (http://srda.sinica.edu.tw/)