You are viewing the RapidMiner Studio documentation for version 7.6 - Check here for latest version
What's New in RapidMiner Studio 7.6.0?
Released: August 15th, 2017
The following describes the enhancements and bug fixes in RapidMiner Studio 7.6.0:
New features
- Sending notification emails can now be configured in the preferences to make use of all modern connection security and authentication mechanisms like TLS 1.2 + PFS
Enhancements
- The sender of notification emails can now be configured in the preferences
- Licenses are now valid for the full last day until midnight
- Improved handling of infeasible parameter values for Self-Organizing Map
- Changed default sampling type parameter for Validation operators to automatic
- Write Message now has a parameter option to append to existing files instead of overwriting them
- Logistic Regression and Generalized Linear Model learners now have a threshold output where they deliver a threshold value optimized for maximal F-measure
- Improved handling of missing and infinite values for Normalize
- Improved handling of missing or broken compatibility numbers in the process xml
- Made behavior of add as label parameter consistent for all cluster operators
- Improved checks for empty example sets in cluster operators
- Improved shown capabilities for cluster operators and added quick fixes for inconsistent parameter selection
- Reduced some internal logging by moving it behind the debug flag which can be activated in the preferences
- Updated Java for Windows and Mac OS X to version 8u141
Bugfixes
- Fixed reproducibility of results when concurrent operators (e.g. Loops) are involved.
- Changing the default connection timeout setting in the preferences now takes effect immediately.
- Sending notification emails now uses the default connection timeout.
- Fixed metadata of Flatten Clustering.
- Fixed behavior of Loop Parameter inside parallel loops.
- Removed unnecessary warning for clustering operators with nominal input data
- Generate Weights (LPR) and Local Polynomial Regression now provide additional kernel parameters for the numerical measure KernelEuclideanDistance instead of failing
- Fixed Gradient Boosted Trees renderer, it no longer shows wrong edge labels and incorrect value sets
- Logistic Regression, Generalized Linear Model, Gradient Boosted Trees and Deep Learning operators no longer crash the software if certain temporary folder permissions are missing
- Logistic Regression and Generalized Linear Model learners now use 0.5 as the threshold as other binominal learners
- Fixed behavior of Loop Attributes when only one attribute is selected for parallel execution
- Fixed Average for Performance inputs that contain AUC
- Fixed side-effects of Apply Threshold in other branches of the process
- Fixed rare crash in Create Association Rules under certain parameter configurations