Visual representation using post modeling feature evaluation

Patent metadata

Field	Value
Publication number	US-12249012-B2
Application number	US-202218056389-A
Country	US
Kind code	B2
Filing date	Nov 17, 2022
Priority date	Nov 17, 2022
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

A method, computer system, and a computer program product are provided for post-modeling feature evaluation. In one embodiment, at least at least one post model visual output and associated data is obtained that at least includes an individual conditional expectation (ICE) plot and a partial dependence (PDP) plot. Using the associated data and the plots, a Feature Importance (PI) plot is provided. A plurality of features is then determined for each PI, PDP and ICE plots to calculate at least one Interesting Value for each plot. An overall score is also calculated for each plurality of features based on the associated Interesting Values for each PDP, ICE and PI plots. At least one top feature is selected based on said scores. A final plot is then generated at least reflecting the top feature. The final plot combines the PI, PDP and ICE plots together.

First claim

What is claimed is:

1. A method for post-modeling feature evaluation, comprising:
obtaining at least one post model visual output and associated data, wherein said visual output includes at least an individual conditional expectation (ICE) plot and a partial dependence (PDP) plot; wherein an ICE plot provides a target value of a particular instance that corresponds to a change in a feature value, and a PDP plot provides a dependency correlation between one or more features on a target value;
using said associated data and said plots, providing a future importance (PI) plot; wherein said PI plot includes a plurality of input values indicating a relative contribution of a dataset on a future generated prediction;
determining a plurality of features for each PI, PDP and ICE plots to calculate at least one Interesting Value for each plot;
determining a turning point (TP) associated with the PDP and ICE plots, wherein said TP is identified by a starting point of a decrease or alternatively an increase on each of said PDP and ICE plots;
calculating an overall score for each plurality of features based and their associated Interesting Values for each PDP, ICE and PI plots and by values associated with the TP;
selecting at least one top feature based on said scores; and
generating a final plot by combining PI, PDP and ICE plots based on said at least one top feature.

2. The method of claim 1 , wherein said features for at least said PDP and ICE plots are determined by extracting a plurality of traits and at least one Turning Point associated with each plot.

3. The method of claim 2 , wherein using said traits include at least a Gradient, an Intercept and a Mean value calculated for each plot.

4. The method of claim 3 , wherein said Interesting Value is calculated based on said calculated traits.

5. The method of claim 3 , wherein each feature for said PI, ICE and PDP plots are normalized reflect an overall priority for each feature.

6. The method of claim 3 , wherein said ICE plot has a plurality of cluster graphs, and each cluster is ranked in order of a priority based on said trait values.

7. The method of claim 6 , wherein said final plot is an ICE plot reflecting each cluster by their ratio and simplified for visualization to show to show at least a particular single trait.

8. A computer system for providing post-modeling feature evaluation, comprising;
one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is enabled to perform the steps:
obtaining at least one post model visual output and associated data, wherein said visual output includes at least an individual conditional expectation (ICE) plot and a partial dependence (PDF) plot; wherein an ICE plot provides a target value of a particular instance that corresponds to a change in a feature value, and a PDP plot provides a dependency correlation between one or more features on a target value;
using said associated data and said plots, providing a future importance (PI) plot; wherein said PI plot includes a plurality of input values indicating a relative contribution of a dataset on a future generated predictions;
determining a plurality of features for each PI, PDP and ICE plots to calculate at least one Interesting Value for each plot;
determining a turning point (TP) associated with the PDP and ICE plots, wherein said TP is identified by a starting point of a decrease or increase on each of said PDP and ICE plots;
calculating an overall score for each plurality of features based and their associated Interesting Values for each PDP, ICE and PI plots and by values associated with the TP;
selecting at least one top feature based on said scores; and
generating a final plot by combining PI, PDP and ICE plots based on said at least one top feature.

9. The computer system of claim 8 , wherein said features for at least said PDP and ICE plots are determined by extracting a plurality of traits and at least one Turning Point associated with each plot.

10. The computer system of claim 9 , wherein using said traits include at least a Gradient, an Intercept and a Mean value calculated for each plot.

11. The computer system of claim 10 , wherein said Interesting Value is calculated based on said calculated traits.

12. The computer system of claim 10 , wherein each feature for said PI, ICE and PDP plots are normalized reflect an overall priority for each feature.

13. The computer system of claim 10 , wherein said ICE plot has a plurality of cluster graphs, and each cluster is ranked in order of a priority based on said trait values.

14. The computer system of claim 13 , wherein said final plot is an ICE plot reflecting each cluster by their ratio and simplified for visualization to show to show at least a particular single trait.

15. A computer program product for providing post-modeling feature evaluation, comprising:
one or more computer-readable storage medium and program instructions stored on at least one of the one or more tangible storage medium, the program instructions executable by a processor, the program instructions comprising:
one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is enabled to perform the steps comprising:
obtaining at least one post model visual output and associated data, wherein said visual output includes at least an individual conditional expectation (ICE) plot and a partial dependence (PDP) plot; wherein an ICE plot provides a target value of a particular instance that corresponds to a change in a feature value, and a PDP plot provides a dependency correlation between one or more features on a target value:

using said associated data and said plots, providing a future importance (PI) plot; wherein said PI plot includes a plurality of input values indicating a relative contribution of a dataset on a future generated predictions;
determining a plurality of features for each PI, PDP and ICE plots to calculate at least one Interesting Value for each plot;
determining a turning point (TP) associated with the PDP and ICE plots, wherein said TP is identified by a starting point of a decrease or increase on each of said PDP and ICE plots;

calculating an overall score for each plurality of features based and their associated Interesting Values for each PDP, ICE and PI plots and by values associated with the TP:
selecting at least one top feature based on said scores; and

generating a final plot by combining PI, PDP and ICE plots based on said at least one top feature.

16. The computer program product of claim 15 , wherein said features for at least said PDP and ICE plots are determined by extracting a plurality of traits and at least one Turning Point associated with each plot.

17. The computer program product of claim 16 , wherein using said traits include at least a Gradient, an Intercept and a Mean value calculated for each plot.

18. The computer program product of claim 17 , wherein said Interesting Value is calculated based on said calculated traits.

19. The computer program p

Assignees

• IBM

Inventors

• Ma Xiao Ming
• Yu wen pei
• Xu Jing James
• Zhang xue ying
• Han Si Er
• Xu Jing
• Wang Jun

Classifications

• G06T11/26 Primary

  Drawing of charts or graphs · CPC title

• G06T11/206 Primary

  Physics · mapped topic