Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@
*/
package org.elasticsearch.xpack.core.ml.dataframe.evaluation.classification;

import org.elasticsearch.common.Nullable;
import org.elasticsearch.common.ParseField;
import org.elasticsearch.common.collect.Tuple;
import org.elasticsearch.common.io.stream.StreamInput;
Expand Down Expand Up @@ -76,25 +75,15 @@ public static Precision fromXContent(XContentParser parser) {
return PARSER.apply(parser, null);
}

private static final int DEFAULT_MAX_CLASSES_CARDINALITY = 1000;
private static final int MAX_CLASSES_CARDINALITY = 1000;

private final int maxClassesCardinality;
private String actualField;
private List<String> topActualClassNames;
private EvaluationMetricResult result;

public Precision() {
this((Integer) null);
}

// Visible for testing
public Precision(@Nullable Integer maxClassesCardinality) {
this.maxClassesCardinality = maxClassesCardinality != null ? maxClassesCardinality : DEFAULT_MAX_CLASSES_CARDINALITY;
}
public Precision() {}

public Precision(StreamInput in) throws IOException {
this.maxClassesCardinality = in.readVInt();
}
public Precision(StreamInput in) throws IOException {}

@Override
public String getWriteableName() {
Expand All @@ -116,7 +105,7 @@ public final Tuple<List<AggregationBuilder>, List<PipelineAggregationBuilder>> a
AggregationBuilders.terms(ACTUAL_CLASSES_NAMES_AGG_NAME)
.field(actualField)
.order(Arrays.asList(BucketOrder.count(false), BucketOrder.key(true)))
.size(maxClassesCardinality)),
.size(MAX_CLASSES_CARDINALITY)),
Collections.emptyList());
}
if (result == null) { // This is step 2
Expand All @@ -142,7 +131,7 @@ public void process(Aggregations aggs) {
if (topActualClassNames == null && aggs.get(ACTUAL_CLASSES_NAMES_AGG_NAME) instanceof Terms) {
Terms topActualClassesAgg = aggs.get(ACTUAL_CLASSES_NAMES_AGG_NAME);
if (topActualClassesAgg.getSumOfOtherDocCounts() > 0) {
// This means there were more than {@code maxClassesCardinality} buckets.
// This means there were more than {@code MAX_CLASSES_CARDINALITY} buckets.
// We cannot calculate average precision accurately, so we fail.
throw ExceptionsHelper.badRequestException(
"Cannot calculate average precision. Cardinality of field [{}] is too high", actualField);
Expand Down Expand Up @@ -175,7 +164,6 @@ public Optional<EvaluationMetricResult> getResult() {

@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeVInt(maxClassesCardinality);
}

@Override
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@
*/
package org.elasticsearch.xpack.core.ml.dataframe.evaluation.classification;

import org.elasticsearch.common.Nullable;
import org.elasticsearch.common.ParseField;
import org.elasticsearch.common.collect.Tuple;
import org.elasticsearch.common.io.stream.StreamInput;
Expand Down Expand Up @@ -70,24 +69,14 @@ public static Recall fromXContent(XContentParser parser) {
return PARSER.apply(parser, null);
}

private static final int DEFAULT_MAX_CLASSES_CARDINALITY = 1000;
private static final int MAX_CLASSES_CARDINALITY = 1000;

private final int maxClassesCardinality;
private String actualField;
private EvaluationMetricResult result;

public Recall() {
this((Integer) null);
}

// Visible for testing
public Recall(@Nullable Integer maxClassesCardinality) {
this.maxClassesCardinality = maxClassesCardinality != null ? maxClassesCardinality : DEFAULT_MAX_CLASSES_CARDINALITY;
}
public Recall() {}

public Recall(StreamInput in) throws IOException {
this.maxClassesCardinality = in.readVInt();
}
public Recall(StreamInput in) throws IOException {}

@Override
public String getWriteableName() {
Expand All @@ -111,7 +100,7 @@ public final Tuple<List<AggregationBuilder>, List<PipelineAggregationBuilder>> a
Arrays.asList(
AggregationBuilders.terms(BY_ACTUAL_CLASS_AGG_NAME)
.field(actualField)
.size(maxClassesCardinality)
.size(MAX_CLASSES_CARDINALITY)
.subAggregation(AggregationBuilders.avg(PER_ACTUAL_CLASS_RECALL_AGG_NAME).script(script))),
Arrays.asList(
PipelineAggregatorBuilders.avgBucket(
Expand All @@ -126,7 +115,7 @@ public void process(Aggregations aggs) {
aggs.get(AVG_RECALL_AGG_NAME) instanceof NumericMetricsAggregation.SingleValue) {
Terms byActualClassAgg = aggs.get(BY_ACTUAL_CLASS_AGG_NAME);
if (byActualClassAgg.getSumOfOtherDocCounts() > 0) {
// This means there were more than {@code maxClassesCardinality} buckets.
// This means there were more than {@code MAX_CLASSES_CARDINALITY} buckets.
// We cannot calculate average recall accurately, so we fail.
throw ExceptionsHelper.badRequestException(
"Cannot calculate average recall. Cardinality of field [{}] is too high", actualField);
Expand All @@ -149,7 +138,6 @@ public Optional<EvaluationMetricResult> getResult() {

@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeVInt(maxClassesCardinality);
}

@Override
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,7 @@ public class ClassificationEvaluationIT extends MlNativeDataFrameAnalyticsIntegT

@Before
public void setup() {
createAnimalsIndex(ANIMALS_DATA_INDEX);
indexAnimalsData(ANIMALS_DATA_INDEX);
}

Expand Down Expand Up @@ -142,12 +143,13 @@ public void testEvaluate_Precision() {
}

public void testEvaluate_Precision_CardinalityTooHigh() {
indexDistinctAnimals(ANIMALS_DATA_INDEX, 1001);
ElasticsearchStatusException e =
expectThrows(
ElasticsearchStatusException.class,
() -> evaluateDataFrame(
ANIMALS_DATA_INDEX,
new Classification(ANIMAL_NAME_FIELD, ANIMAL_NAME_PREDICTION_FIELD, Arrays.asList(new Precision(4)))));
new Classification(ANIMAL_NAME_FIELD, ANIMAL_NAME_PREDICTION_FIELD, Arrays.asList(new Precision()))));
assertThat(e.getMessage(), containsString("Cardinality of field [animal_name] is too high"));
}

Expand All @@ -174,11 +176,12 @@ public void testEvaluate_Recall() {
}

public void testEvaluate_Recall_CardinalityTooHigh() {
indexDistinctAnimals(ANIMALS_DATA_INDEX, 1001);
ElasticsearchStatusException e =
expectThrows(
ElasticsearchStatusException.class,
() -> evaluateDataFrame(
ANIMALS_DATA_INDEX, new Classification(ANIMAL_NAME_FIELD, ANIMAL_NAME_PREDICTION_FIELD, Arrays.asList(new Recall(4)))));
ANIMALS_DATA_INDEX, new Classification(ANIMAL_NAME_FIELD, ANIMAL_NAME_PREDICTION_FIELD, Arrays.asList(new Recall()))));
assertThat(e.getMessage(), containsString("Cardinality of field [animal_name] is too high"));
}

Expand Down Expand Up @@ -283,7 +286,7 @@ public void testEvaluate_ConfusionMatrixMetricWithUserProvidedSize() {
assertThat(confusionMatrixResult.getOtherActualClassCount(), equalTo(2L));
}

private static void indexAnimalsData(String indexName) {
private static void createAnimalsIndex(String indexName) {
client().admin().indices().prepareCreate(indexName)
.addMapping("_doc",
ANIMAL_NAME_FIELD, "type=keyword",
Expand All @@ -293,7 +296,9 @@ private static void indexAnimalsData(String indexName) {
IS_PREDATOR_FIELD, "type=boolean",
IS_PREDATOR_PREDICTION_FIELD, "type=boolean")
.get();
}

private static void indexAnimalsData(String indexName) {
List<String> animalNames = Arrays.asList("dog", "cat", "mouse", "ant", "fox");
BulkRequestBuilder bulkRequestBuilder = client().prepareBulk()
.setRefreshPolicy(WriteRequest.RefreshPolicy.IMMEDIATE);
Expand All @@ -317,4 +322,17 @@ private static void indexAnimalsData(String indexName) {
fail("Failed to index data: " + bulkResponse.buildFailureMessage());
}
}

private static void indexDistinctAnimals(String indexName, int distinctAnimalCount) {
BulkRequestBuilder bulkRequestBuilder = client().prepareBulk()
.setRefreshPolicy(WriteRequest.RefreshPolicy.IMMEDIATE);
for (int i = 0; i < distinctAnimalCount; i++) {
bulkRequestBuilder.add(
new IndexRequest(indexName).source(ANIMAL_NAME_FIELD, "animal_" + i, ANIMAL_NAME_PREDICTION_FIELD, randomAlphaOfLength(5)));
}
BulkResponse bulkResponse = bulkRequestBuilder.get();
if (bulkResponse.hasFailures()) {
fail("Failed to index data: " + bulkResponse.buildFailureMessage());
}
}
}