System and method of using synthetic variables to generate relational Bayesian network models of internet user behaviors

Claim 1

A method for automatically generating a relational Bayesian network to model and predict user web-behavior events in an e-commerce environment using synthetic variables, comprising the acts of: providing a language to represent and create a plurality of synthetic variables; providing an inference using the plurality of synthetic variables; providing a discovery of the plurality of synthetic variables; and wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group consisting of fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain: such input data to a field element in a synthetic variable includes a table on which the field is defined, such as user sessions or user mouse clicks; such input data to a selector element in a synthetic variable includes the values of fields in a database table or output of other elements of the synthetic variable; such input data to a function element of a synthetic variable includes the values of fields in a database table, output of other elements of the synthetic variable, or a subexpression of the synthetic variable; such output data of each field element in a synthetic variable includes data containing the reference or primitive values contained in an appropriate field on the incoming data; such output data of each selector element in a synthetic variable includes a subset of input data to the selector element for which an associated boolean expression evaluates to true; such output data of each function element in a synthetic variable is a value based on arguments to the function; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 2

A method for automatically generating a relational Bayesian network to model and predict user web-behavior events in an e-commerce environment using synthetic variables, comprising the acts of: providing a language to represent and create a plurality of synthetic variables; providing an inference using the plurality of synthetic variables; providing a discovery of the plurality of synthetic variables; and wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group consisting of fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain: such input data to a field element in a synthetic variable includes a table on which the field is defined, such as user sessions or user mouse clicks; such input data to a selector element in a synthetic variable includes the values of fields in a database table or output of other elements of the synthetic variable; such input data to a function element of a synthetic variable includes the values of fields in a database table, output of other elements of the synthetic variable, or a subexpression of the synthetic variable; such output data of each field element in a synthetic variable includes data containing the reference or primitive values contained in an appropriate field on the incoming data; such output data of each selector element in a synthetic variable includes a subset of input data to the selector element for which an associated boolean expression evaluates to true; such output data of each function element in a synthetic variable is a value based on arguments to the function; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 17

A method for searching a database to automatically generate a relational Bayesian network model using synthetic variables comprising: providing a database of information; providing a relational Bayesian network model with the information; providing a language to create from the relational Bayesian network model a plurality of synthetic variables; providing an inference using the plurality of synthetic variables; providing a discovery of the plurality of synthetic variables wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group consisting of fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 18

A method for searching a database to automatically generate a relational Bayesian network model using synthetic variables comprising: providing a database of information; providing a relational Bayesian network model with the information; providing a language to create from the relational Bayesian network model a plurality of synthetic variables; providing an inference using the plurality of synthetic variables; providing a discovery of the plurality of synthetic variables wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group consisting of fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 23

A system for automatically generating a relational Bayesian network model using synthetic variables comprising: a database of information; a language to create from the database of information a plurality of synthetic variables; a processor that provides an inference using the plurality of synthetic variables, wherein the processor further provides a discovery of the plurality of synthetic variables and creates a relational Bayesian network model; and wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group comprising fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 24

A system for automatically generating a relational Bayesian network model using synthetic variables comprising: a database of information; a language to create from the database of information a plurality of synthetic variables; a processor that provides an inference using the plurality of synthetic variables, wherein the processor further provides a discovery of the plurality of synthetic variables and creates a relational Bayesian network model; and wherein: each synthetic variable in the plurality of synthetic variables comprises an expression comprising a chain of one or more elements mapping input data to output data, wherein: each of the one or more elements is selected from the group comprising fields, selectors, functions, variables, and constants; a first element in the chain of one or more elements is a root element that is a name of a database table in a schema; each element in the chain of elements following the first element accepts output data from a previous element and produces input data for a subsequent element; a value of the expression is an output of a last element in the chain; and such input data, such output data, and such synthetic variables are stored in a computer-readable medium.

Claim 29

A method for creating a relational Bayesian network to model and predict a user's real-time behavior in an e-commerce environment comprising the acts of: receiving real-time information regarding a user's first real-time web-behavior; wherein the real-time information is stored in a computer-readable medium; constructing a relational Bayesian network to model and predict a user's second real-time web-behavior, wherein the relational Bayesian network is stored in a computer-readable medium; and modeling the user's second real-time web-behavior by evaluating the Bayesian network, wherein the resulting model is stored in a computer-readable medium.

Claim 30

A method for creating a relational Bayesian network to model and predict a user's real-time behavior in an e-commerce environment comprising the acts of: receiving real-time information regarding a user's first real-time web-behavior; wherein the real-time information is stored in a computer-readable medium; constructing a relational Bayesian network to model and predict a user's second real-time web-behavior, wherein the relational Bayesian network is stored in a computer-readable medium; and modeling the user's second real-time web-behavior by evaluating the Bayesian network, wherein the resulting model is stored in a computer-readable medium.