Core Weka

• Evaluation now computes Matthews correlation coefficient
• AddUserFields filter - A filter that adds new attributes with user specified type and constant value
• Set the configuration of a model corresponding to an entry in the Result list of the Classify, Cluster, Associations or Attribute Selection panel in the Explorer as the current (active) configuration.
• ArffSaver and CSVSaver now have an option to specify how many decimal places to print for numeric attribute values
• CSVLoader now has an option to specify a comma-separated list of characters to use as string enclosures
• GUI and command line package manager now uses GUI proxy authentication dialog from the Bounce framework
• CorrelationAttributeEval attribute evaluator

In Packages

• LibSVM now has a random seed option
• localOutlierFactor - provides the LOF outlier detection algorithm as a filter
• RBFNetwork now includes an RBFClassifier, which trains an RBF network for classification in a fully supervised manner
• AnDE - averaged n dependence estimators from G. Webb's group at Monash (supersedes and replaces the averagedOneDependence estimators package)
• GPAttributeGeneration - genetic programming system to generate attributes. Contributed by Colin Noakes at Loughborough University
• RPlugin - adds integration for the R statistical environment
• SparseGenerativeModel - a wrapper for the Sparse Generative Model toolkit for generative modelling with sparse matrix representations. Comparable to multinomial naive Bayes but with lower time and space complexities
• anonymizationPackage - provides the datafly algorithm for anonymization. Contributed by Matthian Niemann, Daniel Rotar and Maximillian Schroder at the Free University of Berlin