moa.classifiers.meta.imbalanced Tests are failing

[nuwangunasekara]

It looks like moa.classifiers.meta.imbalanced Tests are failing.

How to reproduce

get latest MOA master

(base) ng98@MacBook-Pro-2 moa % git status
On branch master
Your branch is up to date with 'origin/master'.

nothing to commit, working tree clean

(base) ng98@MacBook-Pro-2 moa % git log |head 
commit eeade05db8e929e2ddc1987b2651310c2f76a998
Author: Aljaž Osojnik <aljaz.osojnik@ijs.si>
Date:   Tue May 10 10:40:20 2022 +0200

    Random forest for iSOUP-Tree (#254)

    * iSOUP-Tree integration files

    * Added random forest class for iSOUP-Tree

build MOA using

mvn clean package Attached is the build log: MOA_build_with_test_errors.log

[nuwangunasekara]

Patch 260.txt

• The patch only adds items to ADWIN ensemble in adjustEnsembleSize(), if drift detection is enabled. Drift detection is off by default.

Test logs after the patch After_patch_Testlog.txt

• Failing tests
  • OnlineAdaC2Test
  • OnlineCSB2Test
  • OnlineAdaBoostTest
  • OnlineRUSBoostTest
  • OnlineSMOTEBaggingTest
  • OnlineUnderOverBaggingTest were due to the changes introduced by https://github.com/Waikato/moa/commit/45db18c051044f4dbf8255b0594ced961bf75613 . They are running now. But they have a slight regression in the accuracy.
• Regression in ADOBTest is due to the fix introduced into ADWIN by https://github.com/Waikato/moa/pull/251.
• Above tests are now only failing due to output differences against previous run. Those differences could be due to:
  • https://github.com/Waikato/moa/commit/45db18c051044f4dbf8255b0594ced961bf75613 and
  • https://github.com/Waikato/moa/pull/251

[nuwangunasekara]

Fixed via PR #261 and #263