Make num_top_classes parameter's default value equal to 2

Author: przemekwitek

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/ml/dataframe/analyses/Classification.java

@@ -65,6 +65,11 @@ public static Classification fromXContent(XContentParser parser, boolean ignoreU
         Stream.of(Types.categorical(), Types.discreteNumerical(), Types.bool())
             .flatMap(Set::stream)
             .collect(Collectors.toUnmodifiableSet());
+    /**
+     * As long as we only support binary classification it makes sense to always report both classes with their probabilities.
+     * This way the user can see if the prediction was made with confidence they need.
+     */
+    private static final int DEFAULT_NUM_TOP_CLASSES = 2;
 
     private final String dependentVariable;
     private final BoostedTreeParams boostedTreeParams;
@@ -86,7 +91,7 @@ public Classification(String dependentVariable,
         this.dependentVariable = ExceptionsHelper.requireNonNull(dependentVariable, DEPENDENT_VARIABLE);
         this.boostedTreeParams = ExceptionsHelper.requireNonNull(boostedTreeParams, BoostedTreeParams.NAME);
         this.predictionFieldName = predictionFieldName;
-        this.numTopClasses = numTopClasses == null ? 0 : numTopClasses;
+        this.numTopClasses = numTopClasses == null ? DEFAULT_NUM_TOP_CLASSES : numTopClasses;
         this.trainingPercent = trainingPercent == null ? 100.0 : trainingPercent;
     }
 
@@ -106,6 +111,10 @@ public String getDependentVariable() {
         return dependentVariable;
     }
 
+    public int getNumTopClasses() {
+        return numTopClasses;
+    }
+
     public double getTrainingPercent() {
         return trainingPercent;
     }

x-pack/plugin/core/src/test/java/org/elasticsearch/xpack/core/ml/dataframe/analyses/ClassificationTests.java

@@ -19,6 +19,8 @@
 
 public class ClassificationTests extends AbstractSerializingTestCase<Classification> {
 
+    private static final BoostedTreeParams BOOSTED_TREE_PARAMS = new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0);
+
     @Override
     protected Classification doParseInstance(XContentParser parser) throws IOException {
         return Classification.fromXContent(parser, false);
@@ -43,32 +45,68 @@ protected Writeable.Reader<Classification> instanceReader() {
         return Classification::new;
     }
 
-    public void testConstructor_GivenTrainingPercentIsNull() {
-        Classification classification = new Classification("foo", new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0), "result", 3, null);
-        assertThat(classification.getTrainingPercent(), equalTo(100.0));
-    }
-
-    public void testConstructor_GivenTrainingPercentIsBoundary() {
-        Classification classification = new Classification("foo", new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0), "result", 3, 1.0);
-        assertThat(classification.getTrainingPercent(), equalTo(1.0));
-        classification = new Classification("foo", new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0), "result", 3, 100.0);
-        assertThat(classification.getTrainingPercent(), equalTo(100.0));
-    }
-
     public void testConstructor_GivenTrainingPercentIsLessThanOne() {
         ElasticsearchStatusException e = expectThrows(ElasticsearchStatusException.class,
-            () -> new Classification("foo", new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0), "result", 3, 0.999));
+            () -> new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, 0.999));
 
         assertThat(e.getMessage(), equalTo("[training_percent] must be a double in [1, 100]"));
     }
 
     public void testConstructor_GivenTrainingPercentIsGreaterThan100() {
         ElasticsearchStatusException e = expectThrows(ElasticsearchStatusException.class,
-            () -> new Classification("foo", new BoostedTreeParams(0.0, 0.0, 0.5, 500, 1.0), "result", 3, 100.0001));
+            () -> new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, 100.0001));
 
         assertThat(e.getMessage(), equalTo("[training_percent] must be a double in [1, 100]"));
     }
 
+    public void testConstructor_GivenNumTopClassesIsLessThanZero() {
+        ElasticsearchStatusException e = expectThrows(ElasticsearchStatusException.class,
+            () -> new Classification("foo", BOOSTED_TREE_PARAMS, "result", -1, 1.0));
+
+        assertThat(e.getMessage(), equalTo("[num_top_classes] must be an integer in [0, 1000]"));
+    }
+
+    public void testConstructor_GivenNumTopClassesIsGreaterThan1000() {
+        ElasticsearchStatusException e = expectThrows(ElasticsearchStatusException.class,
+            () -> new Classification("foo", BOOSTED_TREE_PARAMS, "result", 1001, 1.0));
+
+        assertThat(e.getMessage(), equalTo("[num_top_classes] must be an integer in [0, 1000]"));
+    }
+
+    public void testGetNumTopClasses() {
+        Classification classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 7, 1.0);
+        assertThat(classification.getNumTopClasses(), equalTo(7));
+
+        // Boundary condition: num_top_classes == 0
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 0, 1.0);
+        assertThat(classification.getNumTopClasses(), equalTo(0));
+
+        // Boundary condition: num_top_classes == 1000
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 1000, 1.0);
+        assertThat(classification.getNumTopClasses(), equalTo(1000));
+
+        // num_top_classes == null, default applied
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", null, 1.0);
+        assertThat(classification.getNumTopClasses(), equalTo(2));
+    }
+
+    public void testGetTrainingPercent() {
+        Classification classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, 50.0);
+        assertThat(classification.getTrainingPercent(), equalTo(50.0));
+
+        // Boundary condition: training_percent == 1.0
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, 1.0);
+        assertThat(classification.getTrainingPercent(), equalTo(1.0));
+
+        // Boundary condition: training_percent == 100.0
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, 100.0);
+        assertThat(classification.getTrainingPercent(), equalTo(100.0));
+
+        // training_percent == null, default applied
+        classification = new Classification("foo", BOOSTED_TREE_PARAMS, "result", 3, null);
+        assertThat(classification.getTrainingPercent(), equalTo(100.0));
+    }
+
     public void testFieldCardinalityLimitsIsNonNull() {
         assertThat(createTestInstance().getFieldCardinalityLimits(), is(not(nullValue())));
     }

x-pack/plugin/ml/qa/native-multi-node-tests/src/test/java/org/elasticsearch/xpack/ml/integration/ClassificationIT.java

@@ -83,7 +83,7 @@ public void testSingleNumericFeatureAndMixedTrainingAndNonTrainingRows() throws
             assertThat((String) resultsObject.get("keyword-field_prediction"), is(in(KEYWORD_FIELD_VALUES)));
             assertThat(resultsObject.containsKey("is_training"), is(true));
             assertThat(resultsObject.get("is_training"), is(destDoc.containsKey(KEYWORD_FIELD)));
-            assertThat(resultsObject.containsKey("top_classes"), is(false));
+            assertTopClasses(resultsObject, 2, KEYWORD_FIELD, KEYWORD_FIELD_VALUES, String::valueOf);
         }
 
         assertProgress(jobId, 100, 100, 100, 100);
@@ -120,7 +120,7 @@ public void testWithOnlyTrainingRowsAndTrainingPercentIsHundred() throws Excepti
             assertThat((String) resultsObject.get("keyword-field_prediction"), is(in(KEYWORD_FIELD_VALUES)));
             assertThat(resultsObject.containsKey("is_training"), is(true));
             assertThat(resultsObject.get("is_training"), is(true));
-            assertThat(resultsObject.containsKey("top_classes"), is(false));
+            assertTopClasses(resultsObject, 2, KEYWORD_FIELD, KEYWORD_FIELD_VALUES, String::valueOf);
         }
 
         assertProgress(jobId, 100, 100, 100, 100);