What's Changed

• Adding the coefficient of determination and adj. coefficient of determination by @hmgomes in #281
• Updating ARFHoeffdingTree to address issues when subspace size = 1 by @hmgomes in #283
• SelfOptimisingKNearestLeaves and required base learner by @wldbdsyb in #266
• SelfOptimisingKNearestLeaves Test by @wldbdsyb in #286
• Add SGBT Classifier by @nuwangunasekara in #295
• Fixing bug #296 (SGBT does not use evaluation task's random seed) by @nuwangunasekara in #297
• refine and debug the SOKNL and related tree learner by @YibinWaikato in #293
• Fix 294 - Tests on Regressors are failing by @nuwangunasekara in #300
• Add classifiers WeightedkNN and RW_kNN to classifiers.lazy by @marouabahri in #299
• fix: window regression evaluation incremental indcies by @YibinSun in #301
• Fix bug causing HSTreeNode to not use classifierRandom by @justinuliu in #302
• Prediction interval related by @YibinSun in #304
• Yibin: Add HyperPlaneGenerator for Regression by @YibinSun in #305
• Feat: Minibatch bagging classifiers by @cassales in #306
• feat: efficient evaluation loops by @hmgomes in #307
• QueueStream and FilteredQueueStream for CapyMOA pipelines by @heymarco in #308
• Fixing Issue #271 (MOAObject measureByteSize() could return negative value if the size is > 2^31-1) by @nuwangunasekara in #274

New Contributors

• @wldbdsyb made their first contribution in #266
• @YibinWaikato made their first contribution in #293
• @justinuliu made their first contribution in #302
• @cassales made their first contribution in #306
• @heymarco made their first contribution in #308

Full Changelog: 2023.04.0...2024.07.0

MOA 23.04.0

• upgraded fcms-widgets to 0.0.25 to fix bug in directory chooser when using bookmarks (directory tree wasn't expanded properly at the leaf)
• upgraded fcms-widgets to 0.0.24
• switched to SimpleDirectoryChooser for selecting directories (from simple-directory-chooser library)
• upgraded maven-compiler-plugin to 3.10.1; source=target=11 now
• upgraded weka to 3.9.6
• upgraded gson to 2.8.9 to address CVE-2022-25647
• switched scm to git+ssh
• removed jide-oss dependency and look'n'feel (incompatible with Java 17); upgraded jshell-scripting to 0.1.2; upgraded fcms-widgets to 0.0.22; upgraded Meka to 1.9.7
• Setup CD MOA docker workflow (PR #262)
• Test reference files missed in PR #261 and #258 + support to set DJL Engine random seed (PR #263)
• #260 moa.classifiers.meta.imbalanced Tests are failing (PR #261)
• Adding support for CAND (PR #258)
• Random forest for iSOUP-Tree (PR #254)
• New StandardisationFilter (PR #245)
• Update OnlineAdaBoost.java
• Fix resetLearning to clear ArrayLists (PR #250)
• Fixes multilabel classifiers and multilabel generator (PR #249)
• iSOUP-Tree integration files (PR #253)
• Update SAMkNN.java (PR #247)
• #252 IllegalAccessError when trying to browse for an ARFF file in ArffFileStream
• added maven-dependency-plugin to generate GraphML output for all dependency to make tracking down issues in dependencies easier
• Update ADWINChangeDetector.java (PR #251)
• upgraded fcms-widgets to 0.0.21
• upgraded fcms-widgets to 0.0.20
• upgraded jclasslocator to 0.0.19

MOA 2021.07.0

• Added Normalisation, Standardisation, Hashing Trick and Random Projection filters.
• commons.io version bumped from 2.4 to 2.7.
• Added task to write instances to a Kafka topic.
• Last-selected view-mode is now persistent between restarts.

New in this version:

• Regression in kNN.
• New Feature Analysis module.
• New meta-classifiers.
• Meta-task for exporting a task as a Jupyter notebook.

Changes in this version:

• Updated a dependency to fix a bug where the LooknFeel on the Mac platform was throwing an exception.

New in this version:

• Fixed look-and-feel issues on Mac

New in this release:

• New multi-page layout
• New moa-kafka module for reading instances from a Kafka topic
• Adaptive Random Forest defaults now align with original ARF_fast publication
• Addition of the Streaming Random Patches algorithm
• Accuracy-updated ensembles can now use base learners that are not just Hoeffding Trees
• No-Change classifier now available in Lite view mode
• Addition of the ConfStream clustering algorithm
• Bug fixes

Further hotfix for the issue where MOA classes weren't discoverable in Weka.

Dropped the HOTFIX suffix as the Weka package manager doesn't like it.

Hotfix to allow MOA classes to be discovered in Weka using Java11.

New in this release:

• Fixed error when using MOA as a package in WEKA (caused by using the wrong class-loader)
• Fixed error where the class selection dialoque wouldn't set the selected class properly against the parent object.