System and method of using synthetic variables to generate relational Bayesian network models of internet user behaviors
First Claim
Patent Images
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.
3 Assignments
0 Petitions
Accused Products
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.
-
Citations
35 Claims
-
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. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. 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. - View Dependent Claims (17, 18, 19, 20)
-
-
21. 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. - View Dependent Claims (22, 23, 24, 25)
-
-
26. 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. - View Dependent Claims (27, 28, 29, 30, 31, 32, 33, 34, 35)
-
Specification