Granted Patent B2

US 12,724,798 · App. 19/087,745 · Granted Sep 1, 2026

Inventors: Saurabh Batra (Dubai Media City, AE); Dwijendra Dwivedi (Dubai Media City, AE)

Assignee: SAS Institute Inc.

G06F16/287G06F21/6227

Abstract

Claims (168)

1 . A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, the computer-program product including instructions operable to cause a computing system to generate a first feature for a machine learning model by:

receiving identification of character data;

wherein the character data is stored in a database that the computing system is not authorized to access;

wherein the database comprises multiple documents comprising the character data;

wherein one or more documents of the multiple documents pertain to a respective entity of multiple entities; and

wherein the character data comprises identifying information that identifies a given entity when disclosed or misused;

extracting a subset of the character data, wherein the extracting excludes the identifying information;

generating structured data from the subset of the character data by:

generating multiple structured data terms; and

storing the multiple structured data terms in respective data elements of multiple memory blocks in memory of the computing system;

generating the first feature for the machine learning model based on one or more of:

at least two structured data terms in different memory blocks of the multiple memory blocks; and

data manipulation of a structured data term in a memory block of the multiple memory blocks;

generating a virtual, private environment specific to the machine learning model, wherein the virtual, private environment specific to the machine learning model protects included structured data terms used in generating the first feature from users without access to the virtual, private environment specific to the machine learning model; and

generating, using the first feature, the machine learning model in the virtual, private environment specific to the machine learning model.

2 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to:

store the first feature in the memory of the computing system, wherein the memory of the computing system is authorized for access by the computing system and remote from the database;

determine the first feature is a candidate for an independent variable of a machine learning model; and

generate, using the first feature, the machine learning model, wherein the first feature is an independent variable of the machine learning model.

3 . The computer-program product of claim 1 ,

wherein the database comprises a plurality of documents each associated with one of the multiple entities;

wherein the generating the first feature for the machine learning model comprises generating based on at least the data manipulation of the structured data term in the memory block of the multiple memory blocks; and

wherein the instructions are operable to cause the computing system to:

store the first feature in the memory of the computing system, wherein the memory of the computing system is authorized for access by the computing system and remote from the database;

determine the first feature is a predictive indicator for machine learning model to make one or more predictions for the multiple entities; and

generate, using the first feature, the machine learning model for making one or more predictions for the multiple entities.

4 . The computer-program product of claim 1 ,

wherein the database comprises a respective sets of multiple documents, where each respective set of the multiple documents is associated with one of the multiple entities; and

wherein the instructions are operable to cause the computing system to:

store the first feature in the memory of the computing system, wherein the memory of the computing system is authorized for access by the computing system and remote from the database;

determine the first feature is a predictive indicator for machine learning model to make an approval computer-generated recommendation or a disapproval computer-generated recommendation for a new entity that is not a member of the multiple entities;

generate, using the first feature, the machine learning model for making one or more computer-generated recommendations for the multiple entities; and

generate, based on the machine learning model, the approval computer-generated recommendation for the new entity that is not a member of the multiple entities.

5 . The computer-program product of claim 1 ,

wherein the receiving the identification of the character data comprises receiving via a graphical user interface;

wherein the database comprises a respective sets of multiple documents, where each respective set of the multiple documents is associated with one of the multiple entities; and

wherein the instructions are operable to cause the computing system to:

generate, using the first feature, the machine learning model for making one or more predictions for the multiple entities;

generate, based on the machine learning model, a scorecard data model accessing a prediction for a first entity that is not a member of the multiple entities;

store the scorecard data model; and

display, in the graphical user interface, one or more graphical representations of the scorecard data model.

6 . The computer-program product of claim 1 ,

wherein the extracting the subset of the character data comprises:

generating identifiers; and

associating the subset of the character data that is related to a particular entity of the multiple entities with a respective identifier of the identifiers; and

wherein generating the structured data from the subset of the character data comprises:

storing the identifiers in a memory block of the multiple memory blocks; and

associating data in other memory blocks of the multiple data blocks to the identifiers, wherein the associating links data related to a same observation and disassociates data from a particular entity of the multiple entities.

7 . The computer-program product of claim 1 ,

wherein the generating the multiple structured data terms comprises generating headers for respective memory blocks for categorizing the multiple structured data terms; and

wherein generating the first feature comprises:

determining a subset of the headers; and

generating the first feature from the subset of the headers pertaining to the at least two structured data terms in the different memory blocks of the multiple memory blocks.

8 . The computer-program product of claim 1 ,

receiving an adequacy or relevance limitation for data collection; and

wherein the generating the multiple structured data comprises changing granularity of the character data such that the multiple memory blocks complies with the adequacy or relevance limitation for data collection.

9 . The computer-program product of claim 1 ,

wherein the generating the first feature for the machine learning model comprises:

determining there are missing values for structured data; and

generating computer-generated values for the missing values for the structured data; and

wherein the instructions are operable to cause the computing system to:

generate a machine learning model using the first feature and one or more computer-generated values of the computer-generated values.

10 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to:

generate an assessment of the machine learning model; and

display the assessment of the machine learning model in a graphical user interface for user selection of the machine learning model.

11 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to:

obtain a time period;

determine a first set of multiple entities based on the time period;

generate, using the first feature, the machine learning model;

determine a second set of multiple entities based on an updated time period, wherein the first set of multiple entities and the second set of multiple entities comprise different members; and

generate an updated feature or updated machine learning model based on updated ones of the multiple entities according to the updated time period.

12 . The computer-program product of claim 1 ,

wherein one or more documents comprise a plurality of documents;

wherein the plurality of documents comprise at least one document containing character data generated by a first entity; and

wherein the first entity is a member of the multiple entities.

13 . The computer-program product of claim 1 ,

wherein one or more documents comprise a plurality of documents; and

wherein the plurality of documents comprise at least one document containing character data generated about a first entity from a second entity;

wherein the first entity is a member of the multiple entities; and

wherein the second entity is not a member of the multiple entities.

14 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to generate, based on the at least two structured data terms, the first feature for the machine learning model.

15 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to generate the first feature for the machine learning model based on:

a ratio comprising the at least two structured data terms, wherein the ratio obscures individual information for the at least two structured data terms;

an indexing based on the at least two structured data terms, wherein the indexing obscures individual information for the at least two structured data terms; or

a scoring based on the at least two structured data terms, wherein the scoring obscures individual information for the at least two structured data terms.

16 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to generate, based on the data manipulation of a structured data term, the first feature for the machine learning model.

17 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to generate a feature for the machine learning model based on a count, total, or frequency pertaining to a structured data term.

18 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to:

generate the first feature for the machine learning model based on the at least two structured data terms; and

generate a second feature for the machine learning model based on the data manipulation of the structured data term.

19 . The computer-program product of claim 1 , wherein the one or more documents comprise a plurality of documents generated by entities without authorized access to the database, and without authorized access to the computing system.

20 . The computer-program product of claim 1 ,

wherein the receiving the identification of the character data comprises one or more of:

receiving an electronic file containing the character data comprising the identifying information identifying the given entity;

receiving a location of the electronic file containing the character data comprising the identifying information identifying the given entity;

receiving an indication of a selection of the electronic file containing the character data comprising the identifying information identifying the given entity; and

receiving, from the database, the character data comprising the identifying information identifying the given entity.

21 . The computer-program product of claim 1 , wherein extracting the subset of the character data comprises:

refraining from storing, in memory authorized for access by the computing system, the identifying information; or

removing, from the memory authorized for access by the computing system, the identifying information prior to generating the machine learning model.

22 . The computer-program product of claim 1 , wherein generating the first feature comprises:

generating the first feature based on unique data different from the character data; and

storing the unique data in memory that is authorized for access by the computing system, wherein the memory authorized for access by the computing system is devoid of the identifying information.

23 . The computer-program product of claim 1 ,

wherein the identifying information comprises personal data that could identify an identity or contact information for the given entity; and

wherein extracting the subset of the character data comprises identifying, using text mining, the personal data.

24 . The computer-program product of claim 1 ,

wherein the identifying information comprises restricted or confidential information; and

wherein extracting the subset of the character data comprises identifying, using text mining, the restricted or the confidential information.

25 . The computer-program product of claim 1 ,

wherein the instructions are operable to cause the computing system to obtain traits inherent to individuals;

wherein the identifying information comprises the traits; and

wherein extracting a subset of the character data comprises identifying, using computer semantic text processing, text relevant to the trait in the character data.

26 . The computer-program product of claim 1 , wherein the instructions are operable to cause the computing system to:

dynamically allocate, based on text mining of the subset of the character data, at least one of the multiple memory blocks; and

generate the first feature based on one or more structured data terms of the at least one of the multiple memory blocks.

27 . A computer-implemented method comprising generating a first feature for a machine learning model by:

receiving identification of character data;

wherein the character data is stored in a database that a computing system is not authorized to access;

wherein the database comprises multiple documents comprising the character data;

wherein one or more documents of the multiple documents pertain to a respective entity of multiple entities; and

wherein the character data comprises identifying information that identifies a given entity when disclosed or misused;

extracting a subset of the character data, wherein the extracting excludes the identifying information;

generating structured data from the subset of the character data by:

generating multiple structured data terms; and

storing the multiple structured data terms in respective data elements of multiple memory blocks in memory of the computing system; and

generating the first feature for the machine learning model based on one or more of:

at least two structured data terms in different memory blocks of the multiple memory blocks; and

data manipulation of a structured data term in a memory block of the multiple memory blocks;

generating a virtual, private environment specific to the machine learning model, wherein the virtual, private environment specific to the machine learning model protects included structured data terms used in generating the first feature from users without access to the virtual, private environment specific to the machine learning model; and

generating, using the first feature, the machine learning model in the virtual, private environment specific to the machine learning model.

28 . A computing system comprising processor and memory, the memory containing instructions executable by the processor wherein the computing system is configured to generate a first feature for a machine learning model by:

receiving identification of character data;

wherein the character data is stored in a database that the computing system is not authorized to access;

wherein the database comprises multiple documents comprising the character data;

wherein one or more documents of the multiple documents pertain to a respective entity of multiple entities; and

wherein the character data comprises identifying information that identifies a given entity when disclosed or misused;

extracting a subset of the character data, wherein the extracting excludes the identifying information;

generating structured data from the subset of the character data by:

generating multiple structured data terms; and

storing the multiple structured data terms in respective data elements of multiple memory blocks in memory of the computing system;

generating the first feature for the machine learning model based on one or more of:

at least two structured data terms in different memory blocks of the multiple memory blocks; and

data manipulation of a structured data term in a memory block of the multiple memory blocks;

generating a virtual, private environment specific to the machine learning model, wherein the virtual, private environment specific to the machine learning model protects included structured data terms used in generating the first feature from users without access to the virtual, private environment specific to the machine learning model; and

generating, using the first feature, the machine learning model in the virtual, private environment specific to the machine learning model.

29 . A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, the computer-program product including instructions operable to cause a computing system to:

generate a feature for a machine learning model by:

receiving, via a graphical user interface, identification of character data

wherein the character data is stored in a database comprises multiple documents;

wherein one or more documents of the multiple documents pertain to a respective entity of multiple entities;

generating structured data from the character data by:

generating multiple structured data terms; and

storing the multiple structured data terms in respective data elements of multiple memory blocks in memory of the computing system; and

generating the feature for the machine learning model based on one or more of:

at least two structured data terms in different memory blocks of the multiple memory blocks; and

data manipulation of a structured data term in a memory block of the multiple memory blocks;

generate a virtual, private environment specific to the machine learning model, wherein the virtual, private environment specific to the machine learning model protects included structured data terms used in generating the feature from users without access to the virtual, private environment specific to the machine learning model;

generate, using the feature, the machine learning model in the virtual, private environment specific to the machine learning model;

generate, using the feature, the machine learning model for making one or more predictions for the multiple entities;

generate, based on the machine learning model, a scorecard data model for a first entity that is not a member of the multiple entities; and

display, in the graphical user interface, one or more graphical representations of the generated scorecard data model.

30 . The computer-program product of claim 1 ,

wherein the character data comprises raw characters stored within the multiple documents in the database that the computing system is not authorized to access;

wherein extracting the subset of the character data comprises:

identifying a template for the computing system to map to a document of the multiple documents, wherein the template indicates a limitation for collection of any raw characters; and

parsing the document to determine data to restrict in an unstructured format according to the limitation; and

extracting given raw characters in the document in an unstructured format prior to generating the structured data.

Assignments (1)

Continuity (3)

References Cited (20)