Relational Bayesian modeling for electronic commerce
AssigneeArt Technology Group, Inc.
InventorBruce Douglass D'Ambrosio
Life of this patent
Estimating the term…
Abstract
The present invention provides a language, method and system to formulate and evaluate relational Bayesian networks in an e-commerce environment. The present invention employs a specific language for constructing synthetic variables used to predict events in the Bayesian networks. The present system and language allow for efficient and accurate representation, inference, and discovery of the synthetic variables used to model web visitor behavior.
Claims30
A computer-implemented method of modeling one or more events using synthetic variables, the computer-implemented method comprising:
- providing, at a processor, a language to represent a plurality of synthetic variables, wherein the language identifies one or more selection phrases;
- creating, at the processor, the plurality of synthetic variables using the language, wherein each synthetic variable comprises a structured chain of one or more elements, wherein the structured chain acts as a map from input data to output data,
- evaluating, at the processor, the plurality of synthetic variables, wherein each synthetic variable is evaluated to a single value;
- and
- creating, at the processor, one or more models based on the evaluated plurality of synthetic variables.
A system, comprising:
- a processor;
- and
- a computer-readable medium comprising a set of instructions configured to instruct the processor to:
- provide a language to represent a plurality of synthetic variables, wherein the language identifies one or more selection phrases;
- create the plurality of synthetic variables using the language, wherein each synthetic variable comprises a structured chain of one or more elements, wherein the structured chain acts as a map from input data to output data,
- evaluate the plurality of synthetic variables, wherein each synthetic variable is evaluated to a single value;
- and
- create one or more models based on the evaluated plurality of synthetic variables.
A computer-readable medium having instructions stored thereon that, when executed by a processor, cause to processor to model one or more events using synthetic variables, the instructions comprising:
- providing a language to represent a plurality of synthetic variables, wherein the language identifies one or more selection phrases;
- creating the plurality of synthetic variables using the language, wherein each synthetic variable comprises a structured chain of one or more elements, wherein the structured chain acts as a map from input data to output data,
- evaluating the plurality of synthetic variables, wherein each synthetic variable is evaluated to a single value;
- and
- creating one or more models based on the evaluated plurality of synthetic variables.
Dependencies are inferred from the claim text.