|
Activity Number:
|
189
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308360 |
|
Title:
|
Estimating Rates of Rare Events in Massive Web Applications
|
|
Author(s):
|
Deepak Agarwal*+
|
|
Companies:
|
Yahoo! Research
|
|
Address:
|
2825 Mission College Blvd, Santa Clara, CA, 95054,
|
|
Keywords:
|
|
|
Abstract:
|
We consider the problem of estimating occurrence rates of rare events for massive web scale applications. In particular, we focus on the problem of estimating click-through rates for (query, ad) pairs in the context of internet advertising. We consider a scenario where both queries and ads are classified into taxonomies that capture broad contextual information at different levels of granularity. We provide a two stage model that exploits the hierarchical structure induced by the taxonomies and provide accurate estimates at multiple resolutions. Our first stage model is based on a novel imputation strategy that estimates the denominator of the rates. The model also adjusts for the sampling bias introduced due to crawling constraints in our application. Conditional on the first stage estimates, our second stage model uses a tree-structured Markov model to provide smooth rate estimates.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
