What's new in Weka 3.7.8

Core Weka

  • EM and SimpleKMeans now allow for parallel processing on multi cpu/core machines
  • CSVLoader re-written to be more memory efficient and support incremental loading
  • Error plots for classifiers can optionally have point sizes set proportional to the prediction margin
  • Pluggable evaluation metrics for classifiers/regressors
  • Weighted resampling using the Walker's alias method
  • FlowByExpression - KnowledgeFlow component to split incoming instances (or instance stream) according to the evaluation of a logical expression
  • ReplaceMissingWithUserConstant filter
  • PartitionMembership filter - adds partition membership attributes as computed by a classifier that implements PartitionGenerator
  • Stream throughput metrics in the KnowledgeFlow when running incrementally
  • TextSaver KnowledgeFlow component
  • Search facility in the KnowledgeFlow design palette
  • Keyboard shortcuts in the KnowledgeFlow for toolbar buttons
  • Offline mode for the packge manager
  • Improved out of memory detection and new low memory detection for GUIs

In Packages