[ML] Fix DFA feature importance popover empty

x-pack/plugins/ml/public/application/components/data_grid/common.ts

@@ -27,7 +27,11 @@ import {
 
 import { DEFAULT_RESULTS_FIELD } from '../../../../common/constants/data_frame_analytics';
 import { extractErrorMessage } from '../../../../common/util/errors';
-import { FeatureImportance, TopClasses } from '../../../../common/types/feature_importance';
+import {
+  FeatureImportance,
+  FeatureImportanceClassName,
+  TopClasses,
+} from '../../../../common/types/feature_importance';
 
 import {
   BASIC_NUMERICAL_TYPES,
@@ -168,8 +172,9 @@ const getClassName = (className: string, isClassTypeBoolean: boolean) => {
 
   return className;
 };
+
 /**
- * Helper to transform feature importance flattened fields with arrays back to object structure
+ * Helper to transform feature importance fields with arrays back to primitive value
  *
  * @param row - EUI data grid data row
  * @param mlResultsField - Data frame analytics results field
@@ -180,69 +185,44 @@ export const getFeatureImportance = (
   mlResultsField: string,
   isClassTypeBoolean = false
 ): FeatureImportance[] => {
-  const featureNames: string[] | undefined =
-    row[`${mlResultsField}.feature_importance.feature_name`];
-  const classNames: string[] | undefined =
-    row[`${mlResultsField}.feature_importance.classes.class_name`];
-  const classImportance: number[] | undefined =
-    row[`${mlResultsField}.feature_importance.classes.importance`];
-
-  if (featureNames === undefined) {
-    return [];
-  }
-
-  // return object structure for classification job
-  if (classNames !== undefined && classImportance !== undefined) {
-    const overallClassNames = classNames?.slice(0, classNames.length / featureNames.length);
-
-    return featureNames.map((fName, index) => {
-      const offset = overallClassNames.length * index;
-      const featureClassImportance = classImportance.slice(
-        offset,
-        offset + overallClassNames.length
-      );
-      return {
-        feature_name: fName,
-        classes: overallClassNames.map((fClassName, fIndex) => {
+  const featureImportance: Array<{
+    feature_name: string[];
+    classes?: Array<{ class_name: FeatureImportanceClassName[]; importance: number[] }>;
+    importance?: number;
+  }> = row[`${mlResultsField}.feature_importance`];
+  if (featureImportance === undefined) return [];
+
+  return featureImportance.map((fi) => ({
+    feature_name: Array.isArray(fi.feature_name) ? fi.feature_name[0] : fi.feature_name,
+    classes: Array.isArray(fi.classes)
+      ? fi.classes.map((c) => {
+          const processedClass = getProcessedFields(c);
           return {
-            class_name: getClassName(fClassName, isClassTypeBoolean),
-            importance: featureClassImportance[fIndex],
+            importance: processedClass.importance,
+            class_name: getClassName(processedClass.class_name, isClassTypeBoolean),
           };
-        }),
-      };
-    });
-  }
-
-  // return object structure for regression job
-  const importance: number[] = row[`${mlResultsField}.feature_importance.importance`];
-  return featureNames.map((fName, index) => ({
-    feature_name: fName,
-    importance: importance[index],
+        })
+      : fi.classes,
+    importance: Array.isArray(fi.importance) ? fi.importance[0] : fi.importance,
   }));
 };
 
 /**
- * Helper to transforms top classes flattened fields with arrays back to object structure
+ * Helper to transforms top classes fields with arrays back to original primitive value
  *
  * @param row - EUI data grid data row
  * @param mlResultsField - Data frame analytics results field
  * @returns nested object structure of feature importance values
  */
 export const getTopClasses = (row: Record<string, any>, mlResultsField: string): TopClasses => {
-  const classNames: string[] | undefined = row[`${mlResultsField}.top_classes.class_name`];
-  const classProbabilities: number[] | undefined =
-    row[`${mlResultsField}.top_classes.class_probability`];
-  const classScores: number[] | undefined = row[`${mlResultsField}.top_classes.class_score`];
-
-  if (classNames === undefined || classProbabilities === undefined || classScores === undefined) {
-    return [];
-  }
-
-  return classNames.map((className, index) => ({
-    class_name: className,
-    class_probability: classProbabilities[index],
-    class_score: classScores[index],
-  }));
+  const topClasses: Array<{
+    class_name: FeatureImportanceClassName[];
+    class_probability: number[];
+    class_score: number[];
+  }> = row[`${mlResultsField}.top_classes`];
+
+  if (topClasses === undefined) return [];
+  return topClasses.map((tc) => getProcessedFields(tc)) as TopClasses;
 };
 
 export const useRenderCellValue = (

x-pack/test/functional/apps/ml/data_frame_analytics/feature_importance.ts

@@ -14,8 +14,7 @@ export default function ({ getService }: FtrProviderContext) {
   const esArchiver = getService('esArchiver');
   const ml = getService('ml');
 
-  // Failing: See https://github.com/elastic/kibana/issues/90526
-  describe.skip('total feature importance panel and decision path popover', function () {
+  describe('total feature importance panel and decision path popover', function () {
     const testDataList: Array<{
       suiteTitle: string;
       archive: string;