Granted Patent B2

US 10,776,965 · App. 15/493,499 · Granted Sep 15, 2020

Inventors: Robert Chess Stetson (Altadena, CA); Kris Chaisanguanthum (Pasadena, CA); Boris Revechkis (Pasadena, CA); Jacob Aptekar (Pasadena, CA)

Assignee: dRISK, Inc.

G06T11/206G06F16/904G06F16/9024G06F40/137

Abstract

Claims (48)

1. A graph database manipulation device, comprising: a processor; and a memory configured to store a graph database manipulation application; wherein the graph database manipulation application configures the processor to: obtain a tabular data structure, wherein the tabular data structure comprises: a plurality of tables, where each table comprises: a plurality of columns, where each column comprises a header and a plurality of values, and where each unique header in the plurality of columns describes a column type; and a plurality of rows comprising the plurality of values;

generate a graph database from the tabular data structure by configuring the processor to: assign a given node to each column type; assign a particular node to each particular value for each column type in the plurality of columns, such that the particular node is a successor to the given node associated with the column type associated with the particular value indicated by an edge; and connect each particular node to the given node associated with the row associated with the particular value using an edge;

identify a region of interest within a graph described by the graph database;

identify a feature space comprising at least first-order nodes and edges connected to the region of interest;

identify a set of known explanatory variables from the feature space, where known explanatory variables in the set of known explanatory variables comprise nodes and edges within the graph database known to have explanatory power over the region of interest;

calculate a residual subgraph between the feature space and the set of known explanatory variables; and

extract a set of unknown explanatory variables by applying an inference method to the residual subgraph.

2. The device of claim 1 , wherein constructing a feature space further comprises:

integrating first-order connections;

integrating first-order weights;

integrating higher-order connections; and

integrating higher-order weights.

3. The device of claim 1 , wherein extracting the set of unknown explanatory variables from the residual subgraph comprises applying machine learning technique on the residual subgraph.

4. The device of claim 1 , wherein the predictive power of each unknown explanatory variable in the set of unknown explanatory variables is determined using a statistical classifier.

5. The device of claim 1 , wherein the set of unknown explanatory variables is reincorporated into the graph database.

6. The device of claim 1 , wherein the graph database manipulation application further configures the processor to generate at least one supernode.

7. The device of claim 6 , wherein at least one of the at least one supernode is a superfeature comprising data describing at least two features.

8. The device of claim 6 , wherein at least one of the at least one supernode is a superobservation comprising data describing at least two observations.

9. The device of claim 6 , wherein the graph database manipulation application further configures the processor to store the at least one supernode.

10. The device of claim 1 , wherein the graph database manipulation application further configures the processor to:

obtain a hierarchical data structure with attributes; and

convert the hierarchical data structure into a directed acyclic graph with attributes of the hierarchical data structure mapped onto unique nodes in the directed acyclic graph.

11. The device of claim 1 , wherein the inference method is an artificial intelligence method.

12. The device of claim 1 , wherein the graph database manipulation application further configures the processor to generate a directed acyclic graph from the tabular data structure.

13. A method, comprising:

obtaining a tabular data structure, wherein the tabular data structure comprises: a plurality of tables, where each table comprises: a plurality of columns, where each column comprises a header and a plurality of values, and where each unique header in the plurality of columns describes a column type; and a plurality of rows comprising the plurality of values;

generating a graph database using a graph database manipulation device comprising a processor and a memory connected to the processor, wherein the graph database comprises a set of nodes and a set of edges, by: assigning a given node to each column type; assigning a particular node to each particular value for each column type in the plurality of columns, such that the particular node is a successor to the given node associated with the column type associated with the particular value indicated by an edge; and connecting each particular node to the given node associated with the row associated with the particular value using an edge;

identifying a region of interest within a graph described by the graph database using the graph database manipulation device;

identifying a feature space comprising at least first-order and second-order nodes and edges connected to the region of interest using the graph database manipulation device; and

identifying a set of known explanatory variables from the feature space using the graph database manipulation device, where known explanatory variables in the set of known explanatory variables comprise nodes and edges within the graph database known to have explanatory power over the region of interest;

calculating a residual subgraph between the feature space and the set of known explanatory variables; and

extracting a set of unknown explanatory variables by applying an inference method to the residual subgraph.

14. The method of claim 13 , wherein constructing a feature space further comprises:

integrating first-order connections using the graph database manipulation device;

integrating first-order weights using the graph database manipulation device;

integrating higher-order connections using the graph database manipulation device; and

integrating higher-order weights using the graph database manipulation device.

15. The method of claim 13 , further comprising:

obtaining a tabular data structure comprising at least one row and at least one column using the graph database manipulation device; and

converting the tabular data structure into a graph database using the graph database manipulation device.

16. A graph database manipulation device, comprising: a processor; and a memory configured to store a graph database manipulation application; wherein the graph database manipulation application configures the processor to:

obtain a hierarchical data structure, wherein the hierarchical data structure comprises a set of attributes;

generate a graph database from the hierarchical data structure by configuring the processor to convert the hierarchical data structure into a directed acyclic graph with attributes of the hierarchical data structure mapped onto unique nodes in the directed acyclic graph;

identify a region of interest within a graph described by the graph database;

identify a feature space comprising at least first-order nodes and edges connected to the region of interest;

identify a set of known explanatory variables from the feature space, where known explanatory variables in the set of known explanatory variables comprise nodes and edges within the graph database known to have explanatory power over the region of interest;

calculate a residual subgraph between the feature space and the set of known explanatory variables; and

extract a set of unknown explanatory variables by applying an inference method to the residual subgraph.

Assignments (5)

Continuity (5)

Cited By (13)

US 12,197,695 US 12,217,024 US 12,242,821 US 12,271,424 US 12,293,167 US 12,299,416 US 12,314,689 US 12,488,015 US 12,559,136 US 12,602,142 US 12,650,756 US 12,693,769 US 12,713,125