Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[ML] Fixes incorrect feature importance visualization for Data Frame Analytics classification #150816

Merged
merged 7 commits into from
Feb 14, 2023
Merged

[ML] Fixes incorrect feature importance visualization for Data Frame Analytics classification #150816

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
4da12d7
Add fix, add test
qn895 Feb 9, 2023
548f768
Fix row index causing popover to not show up on page >=2
qn895 Feb 10, 2023
08af272
Merge remote-tracking branch 'upstream/main' into dv-fix-classificati…
qn895 Feb 10, 2023
c3f9233
Add tests to check for feature importance when pagination is changed
qn895 Feb 10, 2023
6495166
Merge branch 'main' into dv-fix-classification-path
kibanamachine Feb 13, 2023
bf7b038
Merge branch 'main' into dv-fix-classification-path
kibanamachine Feb 13, 2023
b5a74a4
Merge branch 'main' into dv-fix-classification-path
kibanamachine Feb 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion x-pack/plugins/ml/public/application/components/data_grid/data_grid.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -124,7 +124,7 @@ export const DataGrid: FC<Props> = memo(
analysisType === ANALYSIS_CONFIG_TYPE.OUTLIER_DETECTION
) {
if (schema === 'featureImportance') {
const row = data[rowIndex];
const row = data[rowIndex - pagination.pageIndex * pagination.pageSize];
if (!row) return <div />;
// if resultsField for some reason is not available then use ml
const mlResultsField = resultsField ?? DEFAULT_RESULTS_FIELD;
Expand Down
11 changes: 10 additions & 1 deletion ...ages/analytics_exploration/components/feature_importance/decision_path_classification.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ import type {
} from '../../../../../../../common/types/feature_importance';
import { DecisionPathChart } from './decision_path_chart';
import { MissingDecisionPathCallout } from './missing_decision_path_callout';
import { TopClass } from '../../../../../../../common/types/feature_importance';

interface ClassificationDecisionPathProps {
predictedValue: string | boolean;
Expand All @@ -42,12 +43,20 @@ export const ClassificationDecisionPath: FC<ClassificationDecisionPathProps> = (
const [currentClass, setCurrentClass] = useState<string>(
getStringBasedClassName(topClasses[0].class_name)
);
const selectedClass = topClasses.find(
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The description of the fix refers to True and False base classes, but is this change also the cause of the change in behavior here with the bank_classification_1 MLQA bootstrap job where before I was seeing values greater than 1 for yes and no classes:

Before:

image

With this fix:

image

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

AFAICS, the fix makes total sense here: for the second row, the model predicts the class "no" with the probability of 0.986. This means that the prediction probability of "yes" is 0.014. This is precisely what the screenshot above shows now 🚀

The screenshot before this fix shows that the decision graph briefly exceeds 1.0 in for prediction probability, which is nonsense.

(t) => getStringBasedClassName(t.class_name) === getStringBasedClassName(currentClass)
) as TopClass;
const predictedProbabilityForCurrentClass = selectedClass
? selectedClass.class_probability
: undefined;

const { decisionPathData } = useDecisionPathData({
baseline,
featureImportance,
predictedValue: currentClass,
predictedProbability,
predictedProbability: predictedProbabilityForCurrentClass,
});

const options = useMemo(() => {
const predictionValueStr = getStringBasedClassName(predictedValue);

Expand Down
Loading