INTERACTIVE MACHINE LEARNING MODEL DEVELOPMENT

• US 20200134367A1
• Filed: 10/25/2018
• Published: 04/30/2020
• Est. Priority Date: 10/25/2018
• Status: Active Grant

• Current Assignee: The Boeing Co.
• Original Assignee: The Boeing Co.
• Inventors: Chopra, Seema, Moharir, Akshata Kishore, Sundararaman, Arvind, Kaluskar, Kaustubh
• Granted Patent: US 11,501,103 B2
• Time in Patent Office: Days
• CPC Class Codes: G01C 23/005 Flight directors indicating arrangements specially adapted for rotary gyroscopes G01C19/32
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound

Abstract

A method is provided that includes generating a visual environment for interactive development of a machine learning (ML) model. The method includes accessing observations of data each of which includes values of independent variables and a dependent variable, and performing an interactive exploratory data analysis (EDA) of the values of a set of the independent variables. The method includes performing an interactive feature construction and selection based on the interactive EDA, and in which select independent variables are selected as or transformed into a set of features for use in building a ML model to predict the dependent variable. The method includes building the ML model using a ML algorithm, the set of features, and a training set produced from the set of features and observations of the data. And the method includes outputting the ML model for deployment to predict the dependent variable for additional observations of the data.

Claims

1. An apparatus for interactive machine learning model development, the apparatus comprising:

a memory storing a plurality of observations of data of a system, each of the plurality of observations of the data including values of a plurality of independent variables, and a value of a dependent variable; and

processing circuitry configured to access the memory, and execute an application to generate a visual environment including a graphical user interface (GUI) for interactive development of a machine learning model, according to an iterative process at least an iteration of which includes the apparatus being caused to at least;

access the memory including the plurality of observations of the data;

perform an interactive exploratory data analysis of the values of a set of independent variables from the plurality of independent variables for a set of observations from the plurality of observations of the data, in which infographics are automatically produced in the GUI to visually summarize the values of the set of independent variables, one or more of the set of independent variables or one or more of the set of observations being selected based on user input via the GUI;

perform an interactive feature construction and selection based on the interactive exploratory data analysis, and in which select independent variables from the plurality of independent variables are selected as or transformed into a set of features for use in building the machine learning model to predict the dependent variable, one or more of the select independent variables being selected as or transformed into the set of features based on user input via the GUI and the infographics automatically produced in the GUI; and

build the machine learning model using a machine learning algorithm, the set of features, and a training set produced from the set of features and the plurality of observations of the data, including values of the select independent variables, and the value of the dependent variable,

wherein the interactive development of the machine learning model further includes the apparatus being caused to output the machine learning model for deployment to predict and thereby produce predictions of the dependent variable for additional observations of the data that exclude the value of the dependent variable, the predictions produced by the machine learning model being more accurate than produced by a corresponding machine learning model built without the interactive exploratory data analysis and the interactive feature construction and selection that include user input via the GUI.

2. The apparatus of claim 1, wherein the system is an aircraft, and the plurality of observations of the data is flight data for plurality of flights of the aircraft, for each flight of which the values of the plurality of independent variables are measurements of a plurality of properties recorded by an airborne flight recorder from a plurality of sensors or avionic systems during the flight, and the value of the dependent variable is an indication of a condition of the aircraft during the flight.

3. The apparatus of claim 1, wherein at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

perform the interactive exploratory data analysis in which the set of independent variables or the set of observations is modified based on user input via the GUI, and in which modified infographics are automatically produced in the GUI;

modify one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI and the modified infographics automatically produced in the GUI; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

4. The apparatus of claim 1, wherein at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

modify one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

5. The apparatus of claim 1, wherein the apparatus being caused to perform the interactive feature construction and selection includes being caused to apply the one or more of the select independent variables to a transformation to produce a feature of the set of features, the one or more of the select independent variables or the transformation being selected based on user input via the GUI.

6. The apparatus of claim 5, wherein at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

modify the one or more of the select independent variables or the transformation based on user input via the GUI, and thereby modifying the feature of the set of features produced therefrom, the set of features being a modified set of features including a thereby modified feature for the subsequent iteration; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set.

7. The apparatus of claim 1, wherein the apparatus being caused to build the machine learning model includes being caused to perform an interactive model building in which the machine learning algorithm is selected from a plurality of machine learning algorithms based on user input via the GUI, and wherein at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

perform the interactive model building to build a version of the machine learning model using a different one of the plurality of machine learning algorithms, the set of features, and the training set, the different one of the plurality of machine learning algorithms being selected based on user input via the GUI.

8. The apparatus of claim 1, wherein the iteration of the iterative process further includes the apparatus being caused to:

evaluate the machine learning model, including using the machine learning model to predict and thereby produce evaluative predictions of the dependent variable, and produce at least one evaluative infographic that summarizes the evaluative predictions in a layout that reflects performance of the machine learning model.

9. A method of interactive machine learning model development, the method comprising:

executing an application, via processing circuitry, to generate a visual environment including a graphical user interface (GUI) for interactive development of a machine learning model, according to an iterative process at least an iteration of which includes at least;

accessing a plurality of observations of data of a system, each of the plurality of observations of the data including values of a plurality of independent variables, and a value of a dependent variable;

performing an interactive exploratory data analysis of the values of a set of independent variables from the plurality of independent variables for a set of observations from the plurality of observations of the data, in which infographics are automatically produced in the GUI to visually summarize the values of the set of independent variables, one or more of the set of independent variables or one or more of the set of observations being selected based on user input via the GUI;

performing an interactive feature construction and selection based on the interactive exploratory data analysis, and in which select independent variables from the plurality of independent variables are selected as or transformed into a set of features for use in building the machine learning model to predict the dependent variable, one or more of the select independent variables being selected as or transformed into the set of features based on user input via the GUI and the infographics automatically produced in the GUI; and

building the machine learning model using a machine learning algorithm, the set of features, and a training set produced from the set of features and the plurality of observations of the data, including values of the select independent variables, and the value of the dependent variable; and

outputting the machine learning model for deployment to predict and thereby produce predictions of the dependent variable for additional observations of the data that exclude the value of the dependent variable, the predictions produced by the machine learning model being more accurate than produced by a corresponding machine learning model built without the interactive exploratory data analysis and the interactive feature construction and selection that include user input via the GUI.

10. The method of claim 9, wherein the system is an aircraft, and the plurality of observations of the data is flight data for plurality of flights of the aircraft, for each flight of which the values of the plurality of independent variables are measurements of a plurality of properties recorded by an airborne flight recorder from a plurality of sensors or avionic systems during the flight, and the value of the dependent variable is an indication of a condition of the aircraft during the flight.

11. The method of claim 9, wherein at least one subsequent iteration of the iterative process includes at least:

performing the interactive exploratory data analysis in which the set of independent variables or the set of observations is modified based on user input via the GUI, and in which modified infographics are automatically produced in the GUI;

modifying one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI and the modified infographics automatically produced in the GUI; and

building a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

12. The method of claim 9, wherein at least one subsequent iteration of the iterative process includes at least:

modifying one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI; and

building a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

13. The method of claim 9, wherein performing the interactive feature construction and selection includes applying the one or more of the select independent variables to a transformation to produce a feature of the set of features, the one or more of the select independent variables or the transformation being selected based on user input via the GUI.

14. The method of claim 13, wherein at least one subsequent iteration of the iterative process includes at least:

modifying the one or more of the select independent variables or the transformation based on user input via the GUI, and thereby modifying the feature of the set of features produced therefrom, the set of features being a modified set of features including a thereby modified feature for the subsequent iteration; and

building a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set.

15. The method of claim 9, wherein building the machine learning model includes performing an interactive model building in which the machine learning algorithm is selected from a plurality of machine learning algorithms based on user input via the GUI, and wherein at least one subsequent iteration of the iterative process includes at least:

performing the interactive model building to build a version of the machine learning model using a different one of the plurality of machine learning algorithms, the set of features, and the training set, the different one of the plurality of machine learning algorithms being selected based on user input via the GUI.

16. The method of claim 9, wherein the iteration of the iterative process further includes:

evaluating the machine learning model, including using the machine learning model to predict and thereby produce evaluative predictions of the dependent variable, and producing at least one evaluative infographic that summarizes the evaluative predictions in a layout that reflects performance of the machine learning model.

17. A non-transitory computer-readable storage medium for interactive machine learning model development, the computer-readable storage medium having computer-readable program code stored therein that in response to execution by processing circuitry, causes an apparatus to at least:

generate a visual environment including a graphical user interface (GUI) for interactive development of a machine learning model, according to an iterative process at least an iteration of which includes the apparatus being caused to at least;

access a plurality of observations of data of a system, each of the plurality of observations of the data including values of a plurality of independent variables, and a value of a dependent variable;

perform an exploratory data analysis in which infographics are automatically produced in the GUI to visually summarize the values of a set of independent variables;

perform a feature construction and selection based on the exploratory data analysis, and in which select independent variables are selected as or transformed into a set of features; and

build the machine learning model using a machine learning algorithm, the set of features, and a training set produced from the set of features and the plurality of observations of the data, including values of the select independent variables, and the value of the dependent variable,

wherein one or more of the exploratory data analysis, feature construction and selection, or build of the machine learning model is interactive and based on user input via the GUI, and

wherein the interactive development of the machine learning model further includes the apparatus being caused to output the machine learning model for deployment to predict the dependent variable for additional observations of the data that exclude the value of the dependent variable, predictions produced by the machine learning model being more accurate than produced by a corresponding machine learning model built without the user input via the GUI.

18. The non-transitory computer-readable storage medium of claim 17, wherein the system is an aircraft, and the plurality of observations of the data is flight data for plurality of flights of the aircraft, for each flight of which the values of the plurality of independent variables are measurements of a plurality of properties recorded by an airborne flight recorder from a plurality of sensors or avionic systems during the flight, and the value of the dependent variable is an indication of a condition of the aircraft during the flight.

19. The non-transitory computer-readable storage medium of claim 17, wherein the exploratory data analysis is an interactive exploratory data analysis, and at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

perform the interactive exploratory data analysis in which the set of independent variables or a set of observations for the set of independent variables is modified based on user input via the GUI, and in which modified infographics are automatically produced in the GUI;

modify one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI and the modified infographics automatically produced in the GUI; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

20. The non-transitory computer-readable storage medium of claim 17, wherein the feature construction and selection is an interactive feature construction and selection, and at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

modify one or more of the set of features to produce a modified set of features for the subsequent iteration, based on user input via the GUI; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set produced from the modified set of features and the plurality of observations of the data.

21. The non-transitory computer-readable storage medium of claim 17, wherein the feature construction and selection is an interactive feature construction and selection, and the apparatus being caused to perform the feature construction and selection includes being caused to apply the one or more of the select independent variables to a transformation to produce a feature of the set of features, the one or more of the select independent variables or the transformation being selected based on user input via the GUI.

22. The non-transitory computer-readable storage medium of claim 21, wherein the feature construction and selection is an interactive feature construction and selection, and at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

modify the one or more of the select independent variables or the transformation based on user input via the GUI, and thereby modifying the feature of the set of features produced therefrom, the set of features being a modified set of features including a thereby modified feature for the subsequent iteration; and

build a version of the machine learning model using the machine learning algorithm, the modified set of features, and a modified training set.

23. The non-transitory computer-readable storage medium of claim 17, wherein the apparatus being caused to build the machine learning model includes being caused to perform an interactive model building in which the machine learning algorithm is selected from a plurality of machine learning algorithms based on user input via the GUI, and wherein at least one subsequent iteration of the iterative process includes the apparatus being caused to at least:

perform the interactive model building to build a version of the machine learning model using a different one of the plurality of machine learning algorithms, the set of features, and the training set, the different one of the plurality of machine learning algorithms being selected based on user input via the GUI.

24. The non-transitory computer-readable storage medium of claim 17, wherein the iteration of the iterative process further includes the apparatus being caused to:

evaluate the machine learning model, including using the machine learning model to predict and thereby produce evaluative predictions of the dependent variable, and produce at least one evaluative infographic that summarizes the evaluative predictions in a layout that reflects performance of the machine learning model.