METHOD AND SYSTEM FOR CREATING A PREDICTIVE MODEL FOR TARGETING WEBPAGE TO A SURFER

• US 20100023475A1
• Filed: 07/16/2009
• Published: 01/28/2010
• Est. Priority Date: 07/25/2008
• Status: Active Grant

• Current Assignee: Amedesa, Ltd., LivePerson Incorporated
• Original Assignee: LivePerson Incorporated
• Inventors: Lahav, Shlomo
• Granted Patent: US 8,799,200 B2
• Time in Patent Office: Days
• US Class Current: 706/59
• CPC Class Codes: G06F 16/9577 Optimising the visualization of content, e.g. distillation of HTML documents
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0251 Targeted advertisements

Abstract

A system and method for creating a predictive model to select an object from a group of objects that can be associated with a requested web page, wherein a configuration of the requested web page defines a subgroup of one or more selected objects from the group of objects. Exemplary embodiments of the present invention seek to provide novel solutions for determining which content object, taken from a group of content objects, will be best suited for presentation in association with a link on a web page that has been requested by a certain surfer. Each web page can include one or more links to be associated with content objects from the group.

For each content object presented over a requested web page, a predictive model with relevant predictive factors is processed such that the predicted objective, the probability of success for example, is calculated. A success is defined as a surfer responding to the presented content. For example, should a surfer select a relevant content object, the probability of the objects that can be presented is calculated. Subsequently, the objects with the highest predictive expected utility are selected to be associated with the links in the web page requested by the surfer.

Claims

1. A method for creating a predictive model to select an object from a group of objects that can be associated with a requested web page, wherein a configuration of the requested web page defines a subgroup of one or more selected objects from the group of objects, the method comprising:

collecting predictive information from a plurality of requests to the web-page that were responded by a webpage that included the object and the results of presenting the object;

converting the predictive information into a plurality of predictive factors;

defining a value per each predictive factor;

converting non-ordinal predictive factor values into ordinal predictive factor values;

converting each predictive factor into one or more bins, wherein each bin is associated with a range of predictive factor values and wherein the bins are defined based on their probability of success;

creating a bins table; and

processing the bins table with a predictive model engine configured to generate a predictive model based on data stored in the bins table.

2. The method of claim 1, wherein the step of converting the predictive information into a plurality of predictive factors further comprises converting the configuration of web pages that were transmitted in response to requests into one or more predictive factors.

3. The method of claim 1, wherein the prediction algorithm used by the predictive model engine to generate a predictive model is selected from a group consisting of:

logistic regression, linear regression and decision tree analysis.

4. The method of claim 1, wherein the predictive information includes information associated with a request.

5. The method of claim 4, wherein the associated information has nominal values.

6. The method of claim 4, wherein the associated information represents one or more of:

a surfer'"'"'s IP address;

a receipt time for a request;

a communication application used by a surfer;

or

a URL that is associated with a request.

7. The method of claim 1, wherein the predictive information is embedded within a request and is representative of a surfer'"'"'s past behavior.

8. The method of claim 7, wherein the predictive information is time dependent.

9. The method of claim 7, wherein the predictive information represents one or more of:

time of one or more previous visits by a surfer to a domain that includes a requested webpage;

time that a surfer requested a certain web page;

quantity of a surfer'"'"'s visits to a domain;

or

quantity of selections by a surfer of a certain object.

10. The method of claim 1, wherein the steps of converting each predictive factor into one or more bins and processing the bins table with a predictive model engine configured to generate a predictive model include the use of a property set selected from a group of property sets.

11. The method of claim 10, wherein the steps of converting each predictive factor into one or more bins and processing the bins table with a predictive model engine configured to generate a predictive model are repeated per each property set in the group of property sets for creating a group of predictive models.

12. The method of claim 11, wherein an equivalent predictive model is generated from the group of predictive models.

13. A method for generating a predictive model for selection of an object from a group of objects that can be associated with a requested web page, wherein a configuration of the requested web page defines a subgroup of one or more selected objects from the group of objects, the method comprising:

sampling the data traffic between a plurality surfers and a web server;

storing the sampled data traffic in a temporary memory, wherein each stored data traffic sample comprises;

information retrieved from a request to a web page;

the configuration of the web page that was sent in response to said request; and

the surfer'"'"'s response to said sent web page;

managing one or more pairs of databases, wherein each pair of databases is associated with an object from the group of objects and comprises a success database and a failure database;

per each object within the group of objects, copying from said temporary memory to the success database associated with the object predictive information from the stored data traffic which represents a surfer'"'"'s response that pointed to the object; and

per each object within the group of objects, copying from said temporary memory to the failure database associated with the object predictive information from the stored data traffic which represents a surfer'"'"'s response that did not point to the object.

14. The method of claim 13, wherein a success is indicated by a surfer response which pointed to an object from a subgroup of objects.

15. The method of claim 13, wherein the value of predictive information representing a surfer'"'"'s response that pointed to the object is a function of the frequency of the data traffic samples in which the object was pointed to by the surfer and the value of predictive information representing a surfer'"'"'s response that did not point to the object is a function of the frequency of the data traffic samples in which the object was not pointed by the surfer.

16. The method of claim 13, further comprising:

processing, per each object, the predictive information stored in the object'"'"'s success database and in the object'"'"'s failure database to generate an object'"'"'s prediction model.

17. The method of claim 16, wherein the ratio between the success predictive data and the failure predictive data is used as a weighting factor in creating an object'"'"'s predictive model.