Hi everyone!
New versions of Weka are available for download from the Weka homepage:
* Weka 3.4.18 - the stable book version available as ZIP, Win32 installer, Win32 installer incl. JRE 1.6.0_21 and Mac OS X application.
* Weka 3.6.4 - stable GUI version. It is available as ZIP, with Win32 installer, Win32 installer incl. JRE 1.6.0_21, Win64 installer, Win64 installer incl. 64 bit JRE 1.6.0_21 and Mac OS X application.
* Weka 3.7.3 - development version. It is available as ZIP, with Win32 installer, Win32 installer incl. JRE 1.6.0_21, Win64 installer, Win64 installer incl. 64 bit JRE 1.6.0_21 and Mac OS X application.
All versions contain a significant number of bugfixes, it is recommended to upgrade to the new versions.
Weka homepage:
http://weka.sourceforge.net/
Pentaho Weka micro site:
http://weka.pentaho.com
Pentaho data mining community documentation:
http://wiki.pentaho.com/display/DATA..+Documentation
Packages for Weka>=3.7.2 can be browsed online at:
http://weka.sourceforge.net/packageMetaData/
The Pentaho Weka micro site at http://weka.pentaho.com/ will be updated to reflect the new releases soon.
Note:
It might take a while before Sourceforge.net has propagated all the files to its mirrors.
What's new in 3.7.3?
Some highlights
---------------
* Reduced memory consumption in REPTree. Now allows Leplace-like correction. Bug fixes to the backfitting routine.
* InputMappedClassifier - a new classifier that helps to avoid the dreaded 'train and test set are not compatible' situation by building a mapping between the structure of the data that the base classifier has been trained with and the incoming test data.
* Package manager now checks for new packages at the repository. Now allows 'unofficial' packages to be directly installed from a URL or file in the GUI.
* WekaExcel - a new package that provides loaders and savers for Microsoft Excel spreadsheets.
* WekaODF - a new package that provides loaders and savers for spreadsheets in the open document format.
* classAssociationRules - a new package that provides the ability to use association rule learning algorithms as classifiers. Includes an implementation of the CBA algorithm.
* kfPMMLClassifierScoring - a new package that provides a component for the Knowledge Flow for scoring using PMML classifiers.
* multiInstanceLearning package (version 1.0.1) now has an implementation of the quick diverse density learner (QuickDDIterative).
* Gaussian processes, RegressionByDiscretization and ClassificationViaRegression have moved back into the core distribution from packages.
* ComplementNaiveBayes has moved out of the core distribution into a package called 'complementNaiveBayes.'
* Id3, IB1 and Prism have moved out of the core distribution into a package called 'simpleEducationalLearningSchemes.'
As usual, for a complete list of changes refer to the changelogs.
Cheers,
The Weka Team