Visualize data and significant records based on relationship with the model

Patent metadata

Field	Value
Publication number	US-12293438-B2
Application number	US-202218064959-A
Country	US
Kind code	B2
Filing date	Dec 13, 2022
Priority date	Dec 13, 2022
Publication date	May 6, 2025
Grant date	May 6, 2025

Abstract

In an approach for post-modeling data visualization and analysis, a processor presents a first visualization of a training dataset in a first plot. Responsive to receiving a selection of a data group of the training dataset to analyze, a processor identifies three or fewer key model features of the data group of the training dataset. A processor ascertains a representative record of each key model feature of the three or fewer key model features using a Local Interpretable Model-Agnostic Explanation technique. A processor presents a second visualization of the three or fewer key model features and the representative record of each key model feature in a second plot.

First claim

What is claimed is:

1. A computer-implemented method comprising:
presenting, by one or more processors, a first visualization of a training dataset in a first plot;
responsive to receiving a selection of a data group of the training dataset to analyze, identifying, by the one or more processors, three or fewer key model features of the data group of the training dataset;
ascertaining, by the one or more processors, a representative record of each key model feature of the three or fewer key model features using a Local Interpretable Model-Agnostic Explanation technique;
presenting, by the one or more processors, a second visualization of the three or fewer key model features and the representative record of each key model feature in a second plot;
correcting or completing, by the one or more processors, the training dataset, wherein the training dataset is either incorrect or incomplete;
prior to presenting the first visualization of the training dataset in the first plot, gathering, by one or more processors, the training dataset from one or more sources;
determining, by one or more processors, a degree of importance of the one or more key model features;
ranking, by one or more processors, the one or more key model features according to the degree of importance; and
selecting, by one or more processors, the three or fewer key model features based on a set of criteria, wherein the set of criteria is selected from a group consisting of: a degree of accuracy of each key model feature of the training dataset and a pre-set configuration, further comprises:
selecting, by one or more processors, two key model features of the three or fewer key model features selected; and
condensing, by one or more processors, a key model feature of the three or fewer key model features not selected into a linear combination using a Principle Component Analysis to produce a three-dimension condensed data.

2. The computer-implemented method of claim 1 , further comprising:
subsequent to selecting the three or fewer key model features based on the set of criteria, selecting, by one more processors, a first key model feature from the three or fewer key model features selected;
assigning, by one or more processors, a first value to the first key model feature; and
assigning, by one or more processors, a second value to a second key model feature and a third key model feature of the three or fewer key model features selected.

3. The computer-implemented method of claim 2 , further comprising:
calculating, by one or more processors, the degree of accuracy of the second key model feature and the third key model feature;
determining, by one or more processors, that the degree of accuracy of the second key model feature and the third key model feature exceeds a first threshold; and
responsive to determining the degree of accuracy exceeds the first threshold, designating, by one or more processors, the key model feature as a valid feature.

4. The computer-implemented method of claim 2 , further comprising:
calculating, by one or more processors, the degree of accuracy of the second key model feature;
determining, by one or more processors, that the degree of accuracy of the second key model feature does not exceed the first threshold; and
responsive to determining the degree of accuracy of the second key model feature does not exceed the first threshold, adding, by one or more processors, the second key model feature to a list of candidates.

5. The computer-implemented method of claim 1 , wherein selecting the three or fewer key model features based on the set of criteria further comprises:
clustering, by one or more processors, the training dataset with the three or fewer key model features selected;
comparing, by one or more processors, a first cluster with a baseline; and
selecting, by one or more processors, the three or fewer key model features of the first cluster closest within a second threshold to the baseline.

6. The computer-implemented method of claim 1 , wherein ascertaining the representative record of each key model feature of the three or fewer key model features using a Local Interpretable Model-Agnostic Explanation technique further comprises:
calculating, by one or more processors, a center of a selected cluster;
comparing, by one or more processors, the representative record of each key model feature to the calculated center of the selected cluster;
selecting, by one or more processors, the representative record of the key model feature closest within a fourth threshold to the calculated center of the selected cluster; and
pairing, by one or more processors, the representative record of each key model feature.

7. A computer program product comprising:
one or more computer readable storage media and program instructions stored on the one or more computer readable storage media, the program instructions comprising:
program instructions to present a first visualization of a training dataset in a first plot; responsive to receiving a selection of a data group of the training dataset to analyze, program instructions to identify three or fewer key model features of the data group of the training dataset;
program instructions to ascertain a representative record of each key model feature of the three or fewer key model features using a Local Interpretable Model-Agnostic Explanation technique, further comprises:
program instructions to calculate a center of a selected cluster;
program instructions to compare the representative record of each key model feature to the calculated center of the selected cluster;
program instructions to select the representative record of the key model feature closest within a fourth threshold to the calculated center of the selected cluster; and
program instructions to pair the representative record of each key model feature;

program instructions to present a second visualization of the three or fewer key model features and the representative record of each key model feature in a second plot; and
program instructions to correct or complete the training dataset, wherein the training dataset is either incorrect or incomplete.

8. The computer program product of claim 7 , further comprising:
prior to presenting the first visualization of the training dataset in the first plot, program instructions to gather the training dataset from one or more sources;
program instructions to identify one or more key model features of the training dataset;
program instructions to determine a degree of importance of the one or more key model features;

program instructions to rank the one or more key model features according to the degree of importance; and
program instructions to select the three or fewer key model features based on a set of criteria, wherein the set of criteria is selected from a group consisting of: a degree of accuracy of each key model feature of the training dataset and a pre-set configuration.

9. The computer program product of claim 8 , further comprising:
subsequent to selecting the three or fewer key model features based on the set of criteria, program instructions to select a first key model feature from the three or fewer key model features selected;
program instructions to assign a first value to the first key model feature; and
program instructions to assign a second value to a second key model feature and a third key model feature of the three or fewer key model features selected.

10. The computer program product of claim 9 , further comprising:
program instructions to calculate the degree of accuracy of the second key model feature and the third key model feature;
program instructions to determine that the degree of accuracy of

Assignees

• IBM

Inventors

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

Classifications

• G06T11/26 Primary

  Drawing of charts or graphs · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06T11/206 Primary

  Physics · mapped topic