[backport to 2.x] Add Support for Handling Missing Data in Anomaly Detection (#1274) #1281
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
kaituo
requested review from
jmazanec15,
jngz-es,
saratvemulapalli,
ohltyler,
vamshin,
VijayanB,
ylwu-amzn,
amitgalitz,
jackiehanyang,
sean-zheng-amazon,
dbwiddis,
owaiskazi19 and
joshpalis
as code owners
…h-project#1274) * Add Support for Handling Missing Data in Anomaly Detection This PR introduces enhanced handling of missing data, giving customers the flexibility to choose how to address gaps in their data. Options include ignoring missing data (default behavior), filling with fixed values (customer-specified), zeros, or previous values. These options can improve recall in anomaly detection scenarios. For example, in this forum discussion https://forum.opensearch.org/t/do-missing-buckets-ruin-anomaly-detection/16535, customers can now opt to fill missing values with zeros to maintain detection accuracy. Key Changes: 1. Enhanced Missing Data Handling: Changed to ThresholdedRandomCutForest.process(double[] inputPoint, long timestamp, int[] missingValues) to support missing data in both real-time and historical analyses. The preview mode remains unchanged for efficiency, utilizing existing linear imputation techniques. (See classes: ADColdStart, ModelColdStart, ModelManager, ADBatchTaskRunner). 2. Refactoring Imputation & Processing: Refactored the imputation process, failure handling, statistics collection, and result saving in Inferencer. 3. Improved Imputed Value Reconstruction: Reconstructed imputed values using existing mean and standard deviation, ensuring they are accurately stored in AnomalyResult. Added a featureImputed boolean tag to flag imputed values. (See class: AnomalyResult). 4. Broadcast Support for HC Detectors: Added a broadcast mechanism for HC detectors to identify entity models that haven’t received data in a given interval. This ensures models in memory process all relevant data before imputation begins. Single stream detectors handle this within existing transport messages. (See classes: ADHCImputeTransportAction, ADResultProcessor, ResultProcessor). 5. Introduction of ActionListenerExecutor: Added ActionListenerExecutor to wrap response and failure handlers in an ActionListener, executing them asynchronously using the provided ExecutorService. This allows us to handle responses in the AD thread pool. Testing: Comprehensive testing was conducted, including both integration and unit tests. Of the 7135 lines added and 1683 lines removed, 4926 additions and 749 deletions are in tests, ensuring robust coverage. Signed-off-by: Kaituo Li <[email protected]> * rebase from main Signed-off-by: Kaituo Li <[email protected]> * add comment and remove redundant code Signed-off-by: Kaituo Li <[email protected]> --------- Signed-off-by: Kaituo Li <[email protected]>
amitgalitz
approved these changes
Aug 19, 2024
|
approved, just looks like whitesource security check had issue connecting to the repo |
yep. Thanks. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
backport #1274 to 2.x
Check List
--signoff.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.