Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method and System for Personalized Click Prediction in a Sponsored Search Environment

IP.com Disclosure Number: IPCOM000197202D
Publication Date: 2010-Jun-28
Document File: 3 page(s) / 36K

Publishing Venue

The IP.com Prior Art Database

Related People

Erick Cantu-Paz: INVENTOR [+2]

Abstract

A method and system is provided for personalized click prediction in a sponsored search environment. The method involves developing user-specific features and demography based features. The user-specific features and demography based features are then used in predicting user clicks on advertisements.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 52% of the total text.

Methodand System for Personalized Click Prediction in a Sponsored Search Environment

Abstract

A method and system is provided for personalized click prediction in a sponsored search environment.  The method involves developing user-specific features and demography based features.  The user-specific features and demography based features are then used in predicting user clicks on advertisements.

Description

Disclosed is a method and system for personalized click prediction in a sponsored search environment.

Typically, the probability that users click on ads is predicted in a sponsored search environment and used for influencing ranking, filtering, placement, and pricing of ads.

The method and system disclosed herein, involves developing user-specific features and demography based features that reflect click behavior of individual users and user groups.  The user-specific features and demography based features are developed based on observations of search and click behaviors of a large number of users of a commercial search engine.  Thereafter, the user-specific features and demography based features are integrated into a maximum entropy classification framework and these features contribute to the final predicted clickability score for each query-ad-user tuple.

In an instance, the method and system consider two sets of features, including the demography based features and the user-specific features.  The first set of features, i.e. the demography based features capture the behaviors of group of users segmented based on their demography.  The demographic groups include, but are not limited to, an age based group, a gender based group, a marital status based group, an interest based group, a job status based group, and an occupation based group.  In order to integrate the demography based features into the maximum entropy classification framework, binary features are introduced for each possible value of the demographic groups.  For example, there are eight binary features indicating each of the possible age groups, and only one of these eight features triggers for each user.

Further, in addition to using demographic information in the personalized click prediction, a click feedback feature is also used in capturing historical behavior of user groups.  The click feedback feature is computed subsequent to forming the demographic groups.  For example, historical information for all ads of a specific advertiser accounts is captured using the click feedback feature.  Along the same lines, data may be accumulated for combinations of accounts and demographic groups. ...