Predictive modeling and data analysis in a secure shared system

• US 9,576,246 B2
• Filed: 09/12/2013
• Issued: 02/21/2017
• Est. Priority Date: 10/05/2012
• Status: Active Grant

• Current Assignee: BigML, Inc.
• Original Assignee: BigML, Inc.
• Inventors: Petersen, Poul, Garcia, Beatriz, Blasco, Antonio, Martin, Francisco J., Rovira, Oscar, Verwoerd, Jos, Parker, Charles, Ortega, Jose Antonio, Donaldson, J. Justin, Ashenfelter, Adam
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Nilsson, Eric
• Application Number:
  US14/025,063
• Publication Number:
  US 20140101076A1
• Time in Patent Office: 1,258 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 21/604 Tools and structures for managing or administering access control systems
• CPC Class Codes: G06F 21/6245 Protecting personal data, e.g. for financial or medical purposes
• CPC Class Codes: G06F 2221/2113 Multi-level security, e.g. mandatory access control
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 20/10 specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems

Abstract

A system and method enables users to selectively expose and optionally monetize their data resources, for example on a web site. Data assets such as datasets and models can be exposed by the proprietor on a public gallery for use by others. Fees may be charged, for example, per new model, or per prediction using a model. Users may selectively expose public datasets or public models while keeping their raw data private.

Claims

1. A processor-implemented system comprising:

(a) a source server for managing access to a data source;

(b) a dataset server for creating and managing access to a dataset created from the data source;

(c) a model server for creating and managing access to a model based on the dataset;

(d) a prediction server for creating and managing access to a prediction that results from utilizing the model;

(e) a user interface component that implements a corresponding private dashboard for each one of plural users of the system, wherein each private dashboard stores resources for the corresponding user without exposing the stored resources to any other user of the system; and

(f) a public gallery component that, in cooperation with the user interface, implements a public gallery to enable a first user to selectively advertise a dataset for use by other users of the system to build a model of the dataset, wherein the public gallery component displays a summary of the advertised dataset;

wherein the summary comprises a screen display including, for each field in the dataset, a field name and a histogram of data in the field; and

wherein the public gallery component is configurable to advertise a model for use by other users of the system, by displaying a summary of the advertised model.

2. The system according to claim 1 wherein the system further enables the first user to selectively expose either a white-box public model or a black-box form of a model for use by other users of the system, and wherein the prediction server is configured to make predictions using the black-box form of the model.

3. The system according to claim 1 wherein the system is arranged to clone the dataset advertised in the public gallery to form a copy, and to store the cloned copy in a private dashboard associated with a second user, responsive to an indication of receipt of consideration from the second user.

4. The system according to claim 1 wherein the system further enables the first user to selectively expose a white-box form of a model for use by a second user of the system, wherein the white-box form enables use of the model to make predictions and also discloses a descriptive and actionable version of the model.

5. The system according to claim 4 wherein the system further enables the second user to provide its private dataset to the model server to create a new model accessible in the second user'"'"'s private dashboard.

6. The system according to claim 5 wherein the system further enables the second user to utilize its new model to create a new prediction accessible in the second user'"'"'s private dashboard.

7. The system according to claim 5 wherein the first user elects to charge for each new prediction made by the second user using the exposed model, and the system responsively requires receipt of consideration from the second user for each additional new prediction initiated by the second user.

8. The system of claim 4 wherein the black-box form enables use of the model to make predictions without disclosing its internal operation.

9. The system of claim 4 wherein at least one of the other users of the system is a processor configured to use the system via an application programming interface (API).

10. The system of claim 1 wherein the summary of the advertised dataset comprises metadata that describes the dataset.

11. The system of claim 1 wherein the summary comprises a screen display including, for each field in the dataset, a field name, corresponding type, and a count of records in the dataset that have data in the field.

12. A processor-implemented system comprising:

a dataset server for creating and managing access to a dataset wherein the dataset comprises a plurality of records, each record comprising a plurality of fields;

a model server for creating and managing access to a model based on the dataset, wherein the model is a structured representation of the dataset with predictive power;

a prediction server for creating a prediction from input data by utilizing the model;

a user interface component that implements a corresponding private dashboard for at least one of plural users of the system; and

a public gallery component that implements a public gallery that enables a first user to selectively advertise a dataset in the public gallery for use by other users of the system, wherein the public gallery component displays a summary of the advertised dataset;

wherein the summary of the dataset comprises a screen display including, for each field in the dataset, a field name and a histogram of data in the field; and

wherein the public gallery component further enables the first user to selectively advertise a model owned by the first user for potential use by other users of the system.

13. The system according to claim 12 wherein the system further enables the first user to offer access to the advertised dataset to a second user in exchange for a predetermined consideration.

14. The system according to claim 13 wherein the consideration is monetary.

15. The system according to claim 13 wherein the public gallery component enables the first user to selectively expose either a white-box model or a black-box model for potential use by other users of the system, wherein the black-box form of the model enables use of the model to make predictions without disclosing its internal operation.

16. The system according to claim 12 wherein the system is arranged to clone the dataset advertised in the public gallery to form a copy, and to provide the cloned copy as a private dataset for use by a second user, responsive to an indication of receipt of consideration from the second user.

17. The system of claim 12 wherein the summary of the advertised dataset comprises metadata that describes the dataset.

18. The system of claim 12 wherein the summary comprises a screen display including, for each field in the dataset, a field name, corresponding type, and a count of records in the dataset that have data in the field.

19. The system of claim 12 wherein the public gallery component is configurable to display at least one of a thumbnail picture representation of the model or a number of instances in a dataset used to build the model.

20. A non-transitory, machine readable storage medium having stored thereon a series of instructions for causing one or more processors to perform operations comprising:

accessing a data source to acquire raw data comprising a plurality of records of plural fields;

processing the acquired raw data to form a corresponding dataset;

responsive to input from a first user who owns the dataset, publishing the dataset to a public gallery;

wherein publishing the dataset comprises a screen display including, for each field in the dataset, a field name and a histogram of data in the field;

collecting a fee from a second user to utilize the published dataset;

paying substantially the collected fee to the first user;

cloning the dataset to form a copy; and

providing the copy for exclusive use by the second user;

creating a model of a dataset;

responsive to input from a first user who owns the model, publishing a black-box version of the model to a public gallery, the black-box version enabling use of the model by a second user to make predictions.

21. The non-transitory, machine readable storage medium according to claim 20, the stored instructions further causing the one or more processors to perform operations comprising:

collecting a fee from a second user to utilize the published model to make a new prediction without accessing the internal workings of the model.

22. The non-transitory, machine readable storage medium according to claim 21, the stored instructions further causing the one or more processors to perform operations comprising:

receiving a second dataset from a second user;

conditioned on receiving the said fee, responsive to an input request, cloning the published model from the public gallery to the second user'"'"'s private dashboard;

inputting the second dataset to the cloned model to form a new prediction; and

delivering the new prediction to the second user'"'"'s private dashboard without disclosing internal details of the cloned model; and

crediting at least a part of the fee to the first user.

23. A computer-implemented method comprising:

storing a prediction model in a private workspace associated with a first user of a computing system, so that the prediction model is not visible to any other user of the system;

responsive to an input from the first user, displaying a summary of the prediction model to other users of the system in a public gallery

wherein the summary includes a tree visualization of the prediction model and metadata describing a dataset underlying the prediction model;

receiving input data from a second user of the system;

applying the prediction model to the received input data to form a prediction result; and

providing the prediction result to the second user without disclosing the prediction model.

24. The method of claim 23 wherein applying the prediction model to the received input data is predicated upon receiving a second input from the first user permitting use of the prediction model by the second user.

25. The method of claim 23 including:

in the public gallery, filtering published models to support selection responsive to user input based on one or more variables comprising popularity, recency, white box, black-box, price and or other attributes including names, descriptions, and tags.

26. The method of claim 23 including:

in the public gallery, filtering published datasets to support selection responsive to user input based on one or more variables comprising popularity, recency, number of instances, price and or other attributes including names, descriptions, and tags.