Personalization from incomplete data: what you don't know can hurt

ABSTRACT

Clickstream data collected at any web site (site-centric data) is inherently incomplete, since it does not capture users' browsing behavior across sites (user-centric data). Hence, models learned from such data may be subject to limitations, the nature of which has not been well studied. Understanding the limitations is particularly important since most current personalization techniques are based on site-centric data only. In this paper, we empirically examine the implications of learning from incomplete data in the context of two specific problems: (a) predicting if the remainder of any given session will result in a purchase and (b) predicting if a given user will make a purchase at any future session. For each of these problems we present new algorithms for fast and accurate data preprocessing of clickstream data. Based on a comprehensive experiment on user-level clickstream data gathered from 20,000 users' browsing behavior, we demonstrate that models built on user-centric data outperform models built on site-centric data for both prediction tasks.

REFERENCES

• 1: Gediminas Adomavicius , Alexander Tuzhilin, User profiling in personalization applications through rule discovery and validation, Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining, p.377-381, August 15-18, 1999, San Diego, California, United States [doi>10.1145/312129.312287]
• 2: Aggarwal, C.C., Sun, Z., and Yu, P.S., 1998, Online Generation of Profile Association Rules'. In Proceedings of the Fourth International Conference on Knowledge Discovery and Data Mining.
• 3: Ansari, S., 2000. Integrating E-Commerce and Data Mining: Architecture and Challenges, Web-KDD, Aug., 2000.
• 4: Brodley, C., and Kohavi, R., 2000, Peel the Onion, KDD- CUP 2000, Boston, 2000.
• 5: Chan, P.K., 1999. A Non-Invasive Learning Approach to Building Web User Profiles. In Proceedings WebKDD 1999.
• 6: Matt Cutler, E-metrics: tomorrow's business metrics today (invited talk) (abstract only), Proceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining, p.6, August 20-23, 2000, Boston, Massachusetts, United States [doi>10.1145/347090.347096]
• 7: Johnson, E., Moe, W., Fader, P., Bellman, S., and Lohse, J., 2000, On the Depth and Dynamics of Online Search Behavior, Wharton School Working Paper #00-014, June, 2000.
• 8: Khabaza, T., 2001, "As E-asy as Falling Offa Web Log, Data mining Hits the Web", SPSS Data Mining Magazine, January.
• 9: Kimbrough, S., Padmanabhan, B., and Zheng, Z., 2000, On Usage Metric for Determining Authoritative Sites, In the Proceedings of WITS 2000, Brisbane, Australia.
• 10: Korgaonkar, P., and Wolin, L.D., 1999, A Multivariate analysis of Web usage, J. of Advertising Research, 39, pp 53-68.
• 11: Jesus Mena, Data Mining Your Website, Butterworth-Heinemann, Newton, MA, 1999
• 12: Mobasher, B., Dai H., 2000, Discovery of Aggregate Usage Profiles for Web Personalization, Web-KDD, Aug., 2000
• 13: Mobasher, B., Cooley, R., Srivastava J., 1999, Automatic Personalization Based on Web Usage Mining, Technical Report of Depaul University, TR 99-010.
• 14: Moe, W., and Fader, P., 2000, Which Visits Lead to Purchases? Dynamic Conversion Behavior at e-Commerce Sites, The Wharton School, Working Paper #00-023. Aug. 2000 (A)
• 15: Moe, W., and Fader, P., 2000, Capturing Evolving Visit Behavior in Clickstream Data, The Wharton School, Working Paper #00-003, Aug. 2000 (B).
• 16: Thomas P. Novak , Donna L. Hoffman, New metrics for new media: toward the development of Web measurement standards, World Wide Web Journal, v.2 n.1, p.213-246, Winter 1997
• 17: Nasraoui, O., Frigui, H., Joshi, A., Krishnapuram, R., 1999, Mining Web Access Logs Using Relational Competitive Fuzzy Clustering, In the Proceedings of the Eight International Fuzzy Systems Association World Congress, Taipei, August, 1999.
• 18: Padmanabhan, B., Zheng, Z., Kimbrough, S., 2001, A Comparison of Site-Centric and User-Centric Data Mining Approaches to Predicting Session-Level Purchase Behavior on the Web, The Wharton School OPIM Dept Working Paper 01-01-03.
• 19: Park, Y., Fader, P., 2000, Modeling Browsing Behavior at Multiple Sites, In the Proceedings of Informs Marketing Science Conference, Los Angels, June 2000.
• 20: Perkowitz, M., Etzioni, O, 1997, Adaptive web sites: an AI challenge, In Proceedings of the 15th International Joint Conference on Artificial Intelligence.
• 21: Mike Perkowitz , Oren Etzioni, Adaptive Web sites: automatically synthesizing Web pages, Proceedings of the fifteenth national/tenth conference on Artificial intelligence/Innovative applications of artificial intelligence, p.727-732, July 1998, Madison, Wisconsin, United States
• 22: James E. Pitkow, Summary of WWW characterizations, Computer Networks and ISDN Systems, v.30 n.1-7, p.551-558, April 1, 1998
• 23: Stuart Schechter , Murali Krishnan , Michael D. Smith, Using path profiles to predict HTTP requests, Proceedings of the seventh international conference on World Wide Web 7, p.457-467, April 1998, Brisbane, Australia
• 24: Sen, S., Padmanabhan, B., Tuzhilin, A., White, N., and Stein, R., 1998. The Identification and Satisfaction Of Consumer Analysis-Driven Information Needs Of Marketers on The WWW, European Journal Of Marketing (32:7/8), pp. 688- 702.
• 25: Theusinger, C., Huber, K., 2000. Analyzing the Footsteps of Your Customers, Web-KDD 2000.
• 26: Debra VanderMeer , Kaushik Dutta , Anindya Datta , Krithi Ramamritham , Shamkant B. Navanthe, Enabling scalable online personalization on the Web, Proceedings of the 2nd ACM conference on Electronic commerce, p.185-196, October 17-20, 2000, Minneapolis, Minnesota, United States [doi>10.1145/352871.352892]

INDEX TERMS

Primary Classification:
  H. Information Systems
  H.2 DATABASE MANAGEMENT
      H.2.8 Database applications
          Subjects: Data mining

Additional Classification:
  I. Computing Methodologies
  I.2 ARTIFICIAL INTELLIGENCE
      I.2.6 Learning

Keywords:
Incomplete data, clickstream data, data preprocessing, learning, personalization, probabilistic clipping