Skip to content

Fix change point detection for uncertain non-stationary distributions. #119578

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

Merged
merged 2 commits into from
Jan 7, 2025
Merged

Fix change point detection for uncertain non-stationary distributions. #119578

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5dafe6a
Fix change point detection for uncertain non-stationary distributions.
jan-elastic Jan 6, 2025
732c1de
Replace -1 by ChangeType.NO_CHANGE_POINT
jan-elastic Jan 7, 2025
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
12 changes: 7 additions & 5 deletions ...k/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/aggs/changepoint/ChangeDetector.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,7 +189,7 @@ private TestStats testTrendVs(TestStats H0, double[] values, double[] weights) {
private TestStats testStepChangeVs(TestStats H0, double[] values, double[] weights, int[] candidateChangePoints) {

double vStep = Double.MAX_VALUE;
int changePoint = -1;
int changePoint = ChangeType.NO_CHANGE_POINT;

// Initialize running stats so that they are only missing the individual changepoint values
RunningStats lowerRange = new RunningStats();
Expand Down Expand Up @@ -226,7 +226,7 @@ private TestStats testStepChangeVs(TestStats H0, double[] values, double[] weigh
private TestStats testTrendChangeVs(TestStats H0, double[] values, double[] weights, int[] candidateChangePoints) {

double vChange = Double.MAX_VALUE;
int changePoint = -1;
int changePoint = ChangeType.NO_CHANGE_POINT;

// Initialize running stats so that they are only missing the individual changepoint values
RunningStats lowerRange = new RunningStats();
Expand Down Expand Up @@ -349,7 +349,7 @@ private TestStats testDistributionChange(
) {

double maxDiff = 0.0;
int changePoint = -1;
int changePoint = ChangeType.NO_CHANGE_POINT;

// Initialize running stats so that they are only missing the individual changepoint values
RunningStats lowerRange = new RunningStats();
Expand Down Expand Up @@ -378,10 +378,12 @@ private TestStats testDistributionChange(
// before we run the tests.
SampleData sampleData = sample(values, weights, discoveredChangePoints);
final double[] sampleValues = sampleData.values();
final double[] sampleWeights = sampleData.weights();

double pValue = 1;
for (int cp : sampleData.changePoints()) {
if (cp == ChangeType.NO_CHANGE_POINT) {
continue;
}
double[] x = Arrays.copyOfRange(sampleValues, 0, cp);
double[] y = Arrays.copyOfRange(sampleValues, cp, sampleValues.length);
double statistic = KOLMOGOROV_SMIRNOV_TEST.kolmogorovSmirnovStatistic(x, y);
Expand Down Expand Up @@ -451,7 +453,7 @@ private record TestStats(Type type, double pValue, double var, double nParams, i
}

TestStats(Type type, double pValue, double var, double nParams, DataStats dataStats) {
this(type, pValue, var, nParams, -1, dataStats);
this(type, pValue, var, nParams, ChangeType.NO_CHANGE_POINT, dataStats);
}

boolean accept(double pValueThreshold) {
Expand Down
2 changes: 1 addition & 1 deletion ...n/ml/src/main/java/org/elasticsearch/xpack/ml/aggs/changepoint/ChangePointAggregator.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ public InternalAggregation doReduce(InternalAggregations aggregations, Aggregati
ChangeType change = ChangePointDetector.getChangeType(bucketValues);

ChangePointBucket changePointBucket = null;
if (change.changePoint() >= 0) {
if (change.changePoint() != ChangeType.NO_CHANGE_POINT) {
changePointBucket = extractBucket(bucketsPaths()[0], aggregations, change.changePoint()).map(
b -> new ChangePointBucket(b.getKey(), b.getDocCount(), b.getAggregations())
).orElse(null);
Expand Down
4 changes: 3 additions & 1 deletion x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/aggs/changepoint/ChangeType.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,8 +21,10 @@
*/
public interface ChangeType extends NamedWriteable, NamedXContentObject {

int NO_CHANGE_POINT = -1;

default int changePoint() {
return -1;
return NO_CHANGE_POINT;
}

default double pValue() {
Expand Down
11 changes: 11 additions & 0 deletions ...gin/ml/src/test/java/org/elasticsearch/xpack/ml/aggs/changepoint/ChangeDetectorTests.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@
import java.util.concurrent.atomic.AtomicInteger;
import java.util.stream.DoubleStream;

import static org.hamcrest.Matchers.equalTo;
import static org.hamcrest.Matchers.greaterThan;
import static org.hamcrest.Matchers.instanceOf;
import static org.hamcrest.Matchers.lessThan;
Expand Down Expand Up @@ -243,4 +244,14 @@ public void testProblemDistributionChange() {
ChangeType type = new ChangeDetector(bucketValues).detect(0.05);
assertThat(type, instanceOf(ChangeType.DistributionChange.class));
}

public void testUncertainNonStationary() {
MlAggsHelper.DoubleBucketValues bucketValues = new MlAggsHelper.DoubleBucketValues(
null,
new double[] { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 700, 735, 715 }
);
ChangeType type = new ChangeDetector(bucketValues).detect(0.01);
assertThat(type, instanceOf(ChangeType.NonStationary.class));
assertThat(((ChangeType.NonStationary) type).getTrend(), equalTo("increasing"));
}
}
Loading