Browse Prior Art Database

Machine Learned Models for Selection of Advertisement Providers and Products

IP.com Disclosure Number: IPCOM000239718D
Publication Date: 2014-Nov-27
Document File: 3 page(s) / 134K

Publishing Venue

The IP.com Prior Art Database

Related People

Ming Chang: INVENTOR [+3]

Abstract

A method and system is disclosed for using machine learned models to select advertisement providers and products. The machine learned models assist in determining the advertisements that are required to be shown and the advertisements that will generate maximum revenue.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 55% of the total text.

Machine Learned Models for Selection of Advertisement Providers and Products

Abstract

A method and system is disclosed for using machine learned models to select advertisement providers and products.  The machine learned models assist in determining the advertisements that are required to be shown and the advertisements that will generate maximum revenue.

Description

Disclosed is a method and system for using machine learned models to select advertisement providers and products.  The machine learned models assist in determining the advertisements that are required to be shown and the advertisements that will generate maximum revenue.

The figure below illustrates an exemplary architecture of the system disclosed herein.

Figure

As illustrated, the system includes a model for response prediction and a chooser to determine selection of advertisement providers and products.

The method includes modeling response prediction.  A response rate such as click-through rate (CTR) is predicted for a particular advertisement and associated products.  Traditional prediction models are modeled in the following way:

Such traditional models are modified as:

where prodi is a Boolean variable indicating whether an advertisement product should be enabled for advertisement impression.  Additionally, position bias may be removed and modeled as:

A machine learning model may be trained to predict P.  After training the machine learning model, a chooser may be chosen to determine the advertisement and products to be displayed in a particular advertisement area.  The chooser is restricted by the following constraints:

Local constraints (Intra-ad constraints):Any combination of products within an advertisement...