User interface for machine learning feature engineering studio
A machine learning feature studio comprises a user interface configured to allow a user to define features associated with an entity. The features are calculated using historical or real-time data stored in an event store and associated with the entity. Visualizations and values of the calculated feature are displayed in the user interface and the user may interact with the features, such as to edit and compare them. The user commits the features to the project associated with a machine learning model and selects to export the project. Feature vectors may are calculated using the committed features and are exported to a production environment.
1. A method comprising:
receiving, via one or more user interfaces, one or more user inputs indicating one or more machine learning features associated with event data for one or more entities for use in a machine learning model;
receiving, via the one or more user interfaces, one or more user inputs defining the one or more machine learning features at least by associating selected event data, wherein the selected event data is selected based on the one or more user inputs, with a machine learning feature in the one or more machine learning features and associating at least one operation to perform on the selected event data;
retrieving, from an event store, the selected event data;
generating values for the one or more machine learning features based on the retrieved selected event data and the at least one operation;
generating one or more visualizations of the one or more machine learning features for display on the one or more user interfaces;
generating a feature vector as part of a machine learning model based on the one or more machine learning features; and
exporting the feature vector to a production environment for use in the machine learning model.
2. The method of claim 1 , wherein the generating the one or more machine learning features comprises displaying the event data associated with the one or more entities and receiving a selection of one or more pieces of the data comprising the selected event data;
wherein the values for the one or more machine learning features are generated using the selected one or more pieces of the data.
3. The method of claim 1 , further comprising:
transforming the generated values for the one or more machine learning features based on receiving, via the one or more user interfaces, an indication of a transformation comprising the at least one operation; and
generating one or more visualizations of the transformed values for the one or more machine learning features.
4. The method of claim 1 , further comprising storing the one or more machine learning features as a version.
5. The method of claim 1 , wherein the receiving an indication to export a project comprises receiving an indication of a time; and
wherein the generating the feature vector is based on the time.
6. The method of claim 1 , wherein the visualizations comprise at least one of a bar graph, a scatter plot, a heat map, a pair plot, or other graphical representations.
7. A non-transitory computer-readable medium storing instructions that, when executed, cause operations comprising:
receiving, via one or more user interfaces, one or more user inputs indicating one or more machine learning features associated with event data for one or more entities for use in a machine learning model;
receiving, via the one or more user interfaces, one or more user inputs defining the one or more machine learning features at least by associating selected event data, wherein the selected event data is selected based on the one or more user inputs, with a machine learning feature in the one or more machine learning features and associating at least one operation to perform on the selected event data;
generating values for the, based on the retrieved selected event data and the at least one operation;
generating a feature vector as part of a machine learning model based on the one or more machine learning features; and
exporting the feature vector to a production environment for use in the machine learning model.
8. The non-transitory computer-readable medium of claim 7 , wherein the generating the one or more machine learning features comprises displaying the event data associated with the one or more entities and receiving a selection of one or more pieces of the data comprising the selected data;
wherein the values for the one or more machine learning features are generated using the selected one or more pieces of the data.
9. The non-transitory computer-readable medium of claim 7 , wherein the operations further comprise:
transforming the generated values for the one or more machine learning features based on receiving, via the one or more user interfaces, an indication of a transformation comprising the at least one operation; and
generating one or more visualizations of the transformed values for the one or more machine learning features.
10. The non-transitory computer-readable medium of claim 7 , wherein the operations further comprise storing the one or more machine learning features as a version.
11. The non-transitory computer-readable medium of claim 7 , wherein the receiving an indication to export a project comprises receiving an indication of a time; and
wherein the generating the feature vector is based on the time.
12. The non-transitory computer-readable medium of claim 7 , wherein the operations further comprise generating one or more visualizations of the one or more machine learning features by displaying the one or more visualizations of the one or more machine learning features in a single page.
13. The non-transitory computer-readable medium of claim 12 , wherein the one or more visualizations comprise at least one of a bar graph, a scatter plot, a heat map, a pair plot, or another graphical representations.
14. A feature studio comprising:
an event store configured to store data;
at least one processor;
a computer-readable memory coupled to the at least one processor, the computer-readable memory having stored thereon computer readable instructions that upon execution on the at least one processor causes the feature studio to:
receive, via one or more user interfaces, one or more user inputs indicating one or more machine learning features associated with event data for one or more entities for use in a machine learning model;
receive, via the one or more user interfaces, one or more user inputs defining the one or more machine learning features at least by associating selected event data, wherein the selected event data is selected based on the one or more user inputs, with a machine learning feature in the one or more machine learning features and associating at least one operation to perform on the selected event data;
retrieve, from the event store, the selected event data;
generate, based on the retrieved selected event data and the at least one operation, values for the one or more machine learning features;
generate one or more visualizations of the one or more machine learning features;
generate a feature vector as part of a machine learning model based on the one or more machine learning features; and
export the feature vector to a production environment for use in the machine learning model.
15. The feature studio of claim 14 , wherein the generating the one or more machine learning features comprises displaying the event data associated with the one or more entities and receiving a selection of one or more pieces of the data comprising the selected data;
wherein the values for the one or more machine learning features are generated using the selected one or more pieces of the data.
16. The feature studio of claim 14 , wherein the instructions further cause the feature studio to transform the generated values for the one or more machine learning features based on receiving, via the one or more user interfaces, an indication of a transformation comprising the at least one operation; and
generate one or more visualizations of the transformed values for the one or more machine learning features.
17. The feature studio of claim 14 , wherein the instructions further cause the feature studio to store the one or more machine learning features as a version.
18. The feature studio of claim 14 , wherein the one or more visualizations comprise at least one of a bar graph, a scatter plot, a heat map, a pair plot, or another graphical representations.
19. The feature studio of claim 14 , wherein the receiving the indication to export the project comprises receiving an indication of a time; and
wherein the generating the feature vector is based on the time.
20. The feature studio of claim 14 , wherein the generating the one or more visualizations of the one or more machine learning features comprises displaying the one or more visualizations of the one or more machine learning features in a single page.