Skip to main content

Managing advertising campaigns — an approximate planning approach

Abstract

We consider the problem of displaying commercial advertisements on web pages, in the “cost per click” model. The advertisement server has to learn the appeal of each type of visitor for the different advertisements in order to maximize the profit. Advertisements have constraints such as a certain number of clicks to draw, as well as a lifetime. This problem is thus inherently dynamic, and intimately combines combinatorial and statistical issues. To set the stage, it is also noteworthy that we deal with very rare events of interest, since the base probability of one click is in the order of 10−4. Different approaches may be thought of, ranging from computationally demanding ones (use of Markov decision processes, or stochastic programming) to very fast ones.We introduce NOSEED, an adaptive policy learning algorithm based on a combination of linear programming and multi-arm bandits. We also propose a way to evaluate the extent to which we have to handle the constraints (which is directly related to the computation cost). We investigate the performance of our system through simulations on a realistic model designed with an important commercial web actor.

This is a preview of subscription content, access via your institution.

References

  1. Puterman M L. Markov Decision Processes: Discrete Stochastic Dynamic Programming. New York: John Wiley & Sons, 1994

    MATH  Google Scholar 

  2. Auer P, Cesa-Bianchi N, Fischer P. Finite-time analysis of the multiarmed bandit problem. Machine Learning, 2002, 47(2–3): 235–256

    MATH  Article  Google Scholar 

  3. Abe N, Nakamura A. Learning to optimally schedule Internet banner advertisements. In: Proceedings of the 16th International Conference on Machine Learning. 1999, 12–21

  4. Granmo O C. A Bayesian learning automaton for solving two-armed Bernoulli bandit problems. In: Proceedings of the 7th International Conference on Machine Learning and Applications. 2008, 23–30

  5. Langheinrich M, Nakamura A, Abe N, Kamba T, Koseki Y. Unintrusive customization techniques for web advertising. Computer Networks, 1999, 31(11-16): 1259–1272

    Article  Google Scholar 

  6. Nakamura A, Abe N. Improvements to the linear programming based scheduling of web advertisements. Electronic Commerce Research, 2005, 5(1): 75–98

    MATH  Article  Google Scholar 

  7. Pandey S, Agarwal D, Chakrabarti D, Josifovski V. Bandits for taxonomies: a model-based approach. In: Proceedings of the 7th SIAM International Conference on Data Mining. 2007

  8. Langford J, Zhang T. The epoch-greedy algorithm for multi-armed bandits with side information. In: Proceedings of 20th Advances in Neural Information Processing Systems. 2008, 817–824

  9. Wang C C, Kulkarni S R, Poor H V. Bandit problems with side observations. IEEE Transactions on Automatic Control, 2005, 50(3): 338–355

    MathSciNet  Article  Google Scholar 

  10. Kakade S M, Shalev-Shwartz S, Tewari A. Efficient bandit algorithms for online multiclass prediction. In: Proceedings of the 25th International Conference on Machine Learning. 2008, 440–447

  11. Li W, Wang X, Zhang R, Cui Y, Mao J, Jin R. Exploitation and exploration in a performance based contextual advertising system. In: Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2010, 27–36

  12. Pandey S, Olston C. Handling advertisements of unknown quality in search advertising. In: Proceedings of 18th Advances in Neural Information Processing Systems. 2006, 1065–1072

  13. Agarwal D, Chen B, Elango P. Explore/exploit schemes for web content optimization. In: Proceedings of the 9th IEEE International Conference on Data Mining. 2009, 1–10

  14. Li L, Chu W, Langford J, Schapire R E. A contextual-bandit approach to personalized article recommendation. In: Proceedings of the 19th International Conference on World Wide Web. 2010, 661–670

  15. Richardson M, Dominowska E, Ragno R. Predicting clicks: estimating the click-through rate for new ads. In: Proceedings of the 16th International Conference on World Wide Web. 2007, 521–530

  16. Agarwal D, Chen B C, Elango P. Spatio-temporal models for estimating click-through rate. In: Proceedings of the 18th International Conference on World Wide Web. 2009, 21–30

  17. Agarwal D, Broder A, Chakrabarti D, Diklic D, Josifovski V, Sayyadian M. Estimating rates of rare events at multiple resolutions. In: Proceedings of the 13th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2007, 16–25

  18. Wang X, Li W, Cui Y, Zhang B, Mao J. Clickthrough rate estimation for rare events in online advertising. In: Hua X S, Mei T, Hanjalic A, eds. Online Multimedia Advertising: Techniques and Technologies. Hershey: IGI Global, 2010

    Google Scholar 

  19. Fan T K, Chang C H. Sentiment-oriented contextual advertising. Knowledge and Information Systems, 2010, 23(3): 321–344

    Article  Google Scholar 

  20. Mehta A, Saberi A, Vazirani U, Vazirani V. Adwords and generalized on-line matching. In: Proceedings of the 46th Annual IEEE Symposium on Foundations of Computer Science. 2005, 264–273

  21. Mahdian M, Nazerzadeh H. Allocating online advertisement space with unreliable estimates. In: Proceedings of the 8th ACM Conference on Electronic Commerce. 2007, 288–294

  22. Langford J, Strehl A, Wortman J. Exploration scavenging. In: Proceedings of the 25th International Conference on Machine Learning. 2008, 528–535

  23. Koolen W M, Warmuth M K, Kivinen J. Hedging structured concepts. In: Proceedings of the 23rd Annual Conference on Learning Theory. 2010, 93–105

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Sertan Girgin.

Additional information

Sertan Girgin has two BSc degrees, one in Computer Engineering and the other in Mathematics He also holds a PhD in Computer Engineering from Middle East Technical University (METU), Turkey, 2007. He was a visiting researcher at the Department of Computer Science, University of Calgary, Canada, in 2006. For three years, Dr. Girgin worked as a postdoc researcher in team-project Sequel, INRIA Lille Nord Europe, France. Currently, he is with Google, Inc. His research interests include sequential learning, evolutionary computation, distributed AI and multi-agent systems.

Jérémie Mary is Assistant professor at University of Lille and member of the SequeL team at INRIA. He is also member of the european network of excellence PASCAL 2. He obtained his PhD on online machine learning, at Université Paris XI advised by Michèle Sebag and Antoine Cornuéjols. His main research interests are related to machine learning and more specifically sequential data.With Olivier Nicol (PhD student), he won the ICML’2011 challenge Exploration and Exploitation on data provided by Adobe.

Philippe Preux defended his PhD in Computer Science in 1991, at the Université de Lille, France. He is currently a professor of computer science at the Université de Lille. He is the head of the SequeL research group, affiliated to both INRIA, CNRS, and the university. Since 1991, his research has focused on adaptive systems. He has worked on genetic algorithms and metaheuristics for combinatorial optimization; he then moved to reinforcement learning. These days, his main research interests are statistical learning on sequential data, data mining and sequential decision making in face of very large amounts of data, in non stationary environments.

Olivier Nicol holds a Master’s degree in Computer Science with specialization in software engineering from the University of Lille, France. He is now studying for a PhD under Philippe Preux and Jérémie Mary in the SequeL (Sequential Learning) team at INRIA Lille. His main research interests lie in Machine learning and especially using sequential data such as web logs. For instance he is currently working on how to use data to evaluate recommendation policies (and more generally contextual bndits policies) without having to actually test them on the real world. Together with Jérémy Mary he won the ICML 2011 Exploration and Exploitation challenge which was about balancing exploration and exploitation in order to efficiently recommend items to visitors on an Adobe web site.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Girgin, S., Mary, J., Preux, P. et al. Managing advertising campaigns — an approximate planning approach. Front. Comput. Sci. 6, 209–229 (2012). https://doi.org/10.1007/s11704-012-2873-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11704-012-2873-5

Keywords

  • advertisement selection
  • web sites
  • optimization
  • non-stationary setting
  • linear programming
  • multi-arm bandit
  • click-through rate (CTR) estimation
  • exploration-exploitation trade-off