Record Detail Back
STATISTICAL METHODS IN E-COMMERCE RESEARCH
- Share to:
Internet advertising is a multi-billion-dollar industry, as is evident from the phenomenal success of companies like Google, Yahoo, Microsoft, and continues to grow at a rapid rate. With broadband access becoming ubiquitous, Internet traffic continues to grow in both volume and diversity, providing a rich supply of inventory to be monetized. Fortunately, the surge in supply has been accompanied by an increase in demand, with more dollars being diverted to Internet advertising relative to traditional advertising media like television, radio, and newspapers. Marketplace designs that maximize revenue by exploiting billions of advertising opportunities through efficient allocation of available inventory are the key to success in this scenario. Due to the massive scale of the problem, an attractive way to accomplish this is by learning the statistical behavior of the environment through the huge amounts of data constantly flowing into the system. Furthermore, automated learning reduces overhead and has a low marginal cost per transaction, making Internet advertising a lucrative business. However, learning in these scenarios is highly nontrivial and gives rise to a series of challenging statistical problems, including prediction of rare events from massive amounts of high-dimensional data, experimental designs to learn emerging trends, and protecting advertisers by constantly monitoring traffic quality. In this chapter, I provide a perspective on some of the statistical challenges through illustrative examples.
WOLFGANG JANK AND GALIT SHMUELI - Personal Name
978-0-470-12012-5
NONE
Information Technology
English
2008
1-428
LOADING LIST...
LOADING LIST...