What's New in RapidMiner Studio 7.5.0?

The following describes the enhancements and bug fixes in RapidMiner Studio 7.5.0:

New features

  • The first iteration of new data core that manages data sets in a much more efficient way has arrived! This results in both better performance and less memory usage for the vast majority of operators.
  • Added support for Microsoft Azure Blob Storage with Read Azure Blob Storage, Write Azure Blob Storage, and Loop Azure Blob Storage operators. They work exactly like their existing Amazon S3 counterparts.
  • Added support for Amazon Key Management Service (AWS KMS) for all Amazon S3 operators. You can now optionally add an encryption key id to your Amazon S3 connection to decrypt/encrypt files when working with Amazon S3.
  • Added a new mechanism to provide help, advice messages, and even important announcements to the user.

Enhancements

  • Completely revised result graph interaction, presentation, and visualization (e.g. decision trees, clusters, etc.).
  • It is now possible to highlight the path to a node of a decision tree in the Results view.
  • Cluster nodes in the Results view are now scaled according to their relative size.
  • Undo and redo functionality is now much more intuitive when working with the process canvas. It will now not only restore the process state, but also restore canvas location, operator selection, and the zoom level.
  • Navigating up and down through subprocesses in the UI is now more user friendly. When entering a subprocess and later going back up, you will see the same part of the process you were looking at before entering the subprocess.
  • Remove Duplicates now features a new output port called duplicates which returns the examples identified as duplicates.
  • Fixed memory leaks for Handle Exception, Select Subprocess, and Branch.
  • Execute Script now caches the parsed scripts for significantly faster execution, especially inside Loop operators or other highly concurrent environments. General performance of script execution has also been improved. Also added operator tags and added a default example script to make usage of the operator easier. Last but not least, error messages now include the causing stacktrace for easier debugging.
  • Improved AutoMLP performance.
  • Loading context data shows progress now.
  • Added new global process macro: %{process_start} which captures the timestamp when a process was started.
  • It is now possible to close result tabs with the same shortcut as in your web browser: ctrl+w (command+w on OS X)
  • Added new tutorials for RapidMiner Server and RapidMiner Radoop.
  • Added some more usable date and datetime format defaults to choose from when importing data.
  • Added folder buildingblocks in the .RapidMiner directory which will also be searched for .buildingblock files on startup.
  • The dialog letting you know about an available RapidMiner Studio update now also displays the version number of the update.

Bugfixes

  • Fixed a bug making all parallel Loop operators incredibly resource hungry when running hundreds of thousands of iterations
  • Error bubbles indicating the source of an error in the process now work correctly in nested loops again
  • Removed empty confidence columns when applying the model from Linear Discriminant Analysis, Quadratic Discriminant Analysis, Regularized Discriminant Analysis, Single Rule Induction, Subgroup Discovery
  • Regular Discriminant Analysis no longer ignores the alpha parameter
  • The median for Aggregate now takes the middle point of both middle values in case of an even number of values
  • Fixed error that made operators which use a connection (e.g. Read Salesforce) unusable after importing a process
  • Fixed layout of marketplace search link in operator panel
  • Fixed broken dialog title for package download error
  • Fixed broken configurable entries due to unnecessary escaping
  • Fixed delay when trying to view decision trees in the Results view
  • Fixed major memory leak for Loop, Loop Values, Loop Attributes, and Loop Files
  • Fixed some operator parameter help tooltips being cut off
  • Fixed behaviour of Fast Large Margin if learned with bias (parameter)
  • Fixed pdf/svg image export of the scatter matrix chart
  • Fixed some spelling errors
  • Fixed Linear Regression calculation in case use bias is not selected
  • Fixed confidences of Ada Boost in border cases
  • Logistic Regression and Generalized Linear Model no longer allow p-value calculation without adding intercept
  • Fixed problem when trying to delete extensions of which more than one version was installed

Developers

  • Concurrency API introduced with 7.4.0 is now available for unsigned extensions

Notes

  • Changes to Fast Large Margin might affect behaviour of models learned with prior versions of RapidMiner. If you have an existing Fast Large Margin model which was learned using bias, we suggest you learn the model again with this release to ensure correct predictions.