SYSTEMS AND METHODS OF DATA ANALYTICS

• US 20140337320A1
• Filed: 05/13/2013
• Published: 11/13/2014
• Est. Priority Date: 05/13/2013
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Hernandez, Andres Quiroz, Vandervort, David R., Kataria, Saurabh
• Granted Patent: US 9,378,250 B2
• Time in Patent Office: Days
• US Class Current: 707/722
• CPC Class Codes: G06F 16/2428 Query predicate definition using graphical user interfaces, including menus and forms G06F16/2423 takes precedence
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases

Abstract

Systems and methods of data analytics, which in various embodiments enable business analysts to apply certain machine learning and analytics algorithms in a self-service manner by binding them to generic business questions that they can be used to answer in particular domains. The general approach may be to define the application of an algorithm to solve specific problems (questions) for particular combinations of a business domain and a data category. At design time, the algorithm may be linked to canonical data within a data category and programmed to run with this canonical data set. At runtime, given a dataset and its category, and a business domain, a user may choose from the corresponding questions and the system may run the algorithm bound to that question.

Claims

1. A method of performing data analytics, the method being performed on a data analytics system comprising a non-transitory computer-readable storage medium and a processor attached thereto, the method comprising:

storing, in the computer-readable storage, one or more applications, each application being associated with an algorithm, each application being further associated with canonical data indicative of a class of data to be accepted by the algorithm associated with the application;

storing, in the computer-readable storage, one or more questions, each question being associated with an application;

storing a user dataset associated with a domain and a data category;

selecting a question from the one or more questions, the selected question being selected based at least in part on the domain and the data category of the user dataset;

matching the user dataset based on the canonical data of the application associated with the selected question, the matching being performed by the processor, the matching comprising comparing one or more fields of the user dataset with the class of data indicated by the canonical data, the matching thereby producing a canonicalized dataset;

executing the algorithm associated with the application, wherein the canonicalized dataset is provided as input to the algorithm; and

presenting output from the algorithm to the user.

2. The method of claim 1, wherein each question further comprises a domain and a user category, and wherein selecting the question from the one or more questions comprises identifying a subset of matching questions having the same domain and user category as the domain and the user category of the user dataset.

3. The method of claim 2, wherein the one or more questions include a basic question, and wherein selecting the question from the one or more questions further comprises selecting the basic question upon a determination that no question matches the domain and user category of the user dataset.

4. The method of claim 3, wherein the basic question includes a question text containing an interpolation flag, and wherein selecting the basic question comprises interpolating user-provided text with the question text of the basic question to produce interpolated question text, and presenting the interpolated question text to a user.

5. The method of claim 1, wherein the canonical data identifies one or more data field descriptors, and wherein matching the user dataset based on the canonical data comprises selecting fields from the user dataset based at least in part on the one or more data field descriptors of the canonical data.

6. The method of claim 1, wherein each application further comprises data indicative of a mapping to the algorithm of the application, and wherein executing the algorithm associated with the application comprises mapping the canonicalized data to the input to the algorithm based on the data indicative of the mapping.

7. The method of claim 1, wherein the user dataset is received from an external computer system via a network interface of the analytics system.

8. The method of claim 7, further comprising:

identifying a subset of matching questions from the one or more questions based at least in part on the domain and data category of the user dataset;

transmitting, to the external computer system, a user interface identifying the subset of matching questions; and

receiving a user form response from the external computer system via the transmitted user interface, wherein selecting the question from the one or more questions is based at least in part on the user form response.

9. The method of claim 1, wherein the canonical data for each application is determined based on an abstraction operation performed by the analytics system, the abstraction operation comprising:

identifying an example dataset associated with the application; and

determining one or more field descriptors based on the example dataset.

10. The method of claim 9, wherein determining the one or more field descriptors is based at least in part on a selection of fields by an operator of the analytics system.

11. The method of claim 9, wherein each field descriptor identifies a data type and one or more characteristics of a data field.

12. The method of claim 9, wherein the one or more field descriptors are determined based at least in part on inputs associated with the algorithm of the application.

13. The method of claim 9, wherein the example dataset is uploaded to the analytics system by an operator of the analytics system.

14. The method of claim 1, wherein matching the user dataset based on the canonical data comprises identifying an ambiguity in matching at least one descriptor of the canonical data, and requesting manual interaction to resolve the ambiguity.

15. The method of claim 14, wherein the ambiguity involves a descriptor of the canonical data not matching any field of the user dataset or matching multiple fields of the user dataset.

16. A computer system, comprising:

an application store comprising computer-readable storage, having stored therein one or more applications, each application being associated with an algorithm, each application being further associated with canonical data indicative of a class of data to be accepted by the algorithm associated with the application;

a question store comprising computer-readable storage, having stored therein one or more questions, each question being associated with an application;

a user dataset store comprising computer-readable storage, having stored therein a user dataset associated with a domain and a data category;

a question selection module configured to select a question from the one or more questions, the selected question being selected based at least in part on the domain and the data category of the user dataset;

a dataset matching module configured to match the user dataset based on the canonical data of the application associated with the selected question, the matching being performed by the processor, the matching comprising comparing one or more fields of the user dataset with the class of data indicated by the canonical data, the matching thereby producing a canonicalized dataset; and

an application execution module configured to execute the algorithm associated with the application, wherein the canonicalized dataset is provided as input to the algorithm.

17. A method of performing data analytics, the method being performed using a computer processor, the method comprising:

receiving a user dataset;

selecting a question that may be answered with respect to the user dataset from a plurality of questions, the selection being based on stored attributes of the plurality of questions and further being based on attributes of the user dataset;

reconfiguring the user dataset to conform with one or more inputs associated with an algorithm, the algorithm being identified by the computer processor as being configured to respond to the selected question; and

executing the algorithm based on the reconfigured user dataset.