[7.x][ML] Ensure class is represented when its cardinality is low (#56783) (#56829)

In DF analytics classification, it is possible to use no samples
of a class if its cardinality is too low.

This commit fixes this by ensuring the target sample count can never be zero.

Backport of #56783

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/dataframe/process/crossvalidation/StratifiedCrossValidationSplitter.java

@@ -57,9 +57,13 @@ public class StratifiedCrossValidationSplitter implements CrossValidationSplitte
             throw new IllegalStateException("Unknown class [" + classValue + "]; expected one of " + classSamples.keySet());
         }
 
+        // We ensure the target sample count is at least 1 as if the cardinality
+        // is too low we might get a target of zero and, thus, no samples of the whole class
+        double targetSampleCount = Math.max(1.0, samplingRatio * sample.cardinality);
+
         // The idea here is that the probability increases as the chances we have to get the target proportion
         // for a class decreases.
-        double p = (samplingRatio * sample.cardinality - sample.training) / (sample.cardinality - sample.observed);
+        double p = (targetSampleCount - sample.training) / (sample.cardinality - sample.observed);
 
         boolean isTraining = random.nextDouble() <= p;
 

x-pack/plugin/ml/src/test/java/org/elasticsearch/xpack/ml/dataframe/process/crossvalidation/StratifiedCrossValidationSplitterTests.java

@@ -225,6 +225,35 @@ public class StratifiedCrossValidationSplitterTests extends ESTestCase {
         }
     }
 
+    public void testProcess_GivenTwoClassesWithCardinalityEqualToOne_ShouldUseForTraining() {
+        dependentVariable = "dep_var";
+        fields = Arrays.asList(dependentVariable, "feature");
+        classCardinalities = new HashMap<>();
+        classCardinalities.put("class_a", 1L);
+        classCardinalities.put("class_b", 1L);
+        CrossValidationSplitter splitter = createSplitter(80.0);
+
+        {
+            String[] row = new String[]{"class_a", "42.0"};
+
+            String[] processedRow = Arrays.copyOf(row, row.length);
+            splitter.process(processedRow, this::incrementTrainingDocsCount, this::incrementTestDocsCount);
+
+            assertThat(Arrays.equals(processedRow, row), is(true));
+        }
+        {
+            String[] row = new String[]{"class_b", "42.0"};
+
+            String[] processedRow = Arrays.copyOf(row, row.length);
+            splitter.process(processedRow, this::incrementTrainingDocsCount, this::incrementTestDocsCount);
+
+            assertThat(Arrays.equals(processedRow, row), is(true));
+        }
+
+        assertThat(trainingDocsCount, equalTo(2L));
+        assertThat(testDocsCount, equalTo(0L));
+    }
+
     private CrossValidationSplitter createSplitter(double trainingPercent) {
         return new StratifiedCrossValidationSplitter(fields, dependentVariable, classCardinalities, trainingPercent, randomizeSeed);
     }