You are viewing the RapidMiner Studio documentation for version 9.8 - Check here for latest version
What's New in RapidMiner Studio 9.5.0?
Released: Nov 05th, 2019
The following describes the bug fixes in RapidMiner Studio 9.5.0:
New Features
- Added ability to upgrade RapidMiner Studio independently from Server. You can now connect to and access data and processes on older Server versions (9.0 and above) with any current or future Studio version! Processes and data are stored as-is on Server, which enables effective collaboration with your colleagues. However, you need to be aware that while you are able to store processes with brand-new operators on older Servers, you obviously can only run processes that consist of operators that the old Server knows about.
- Deployments: Deployments can now be copied from one to another deployment location (for example from a test to a production server)
Enhancements
- Improved performance of Principal Component Analysis and Weight by PCA operators
- The Import Data dialog now detects files with non-lowercase file extensions
- Fixed view order for Deployments view
- Visualizations: Fixed various issues that could cause Studio startup to fail
- Visualizations: Fixed various issues that could cause them to not be displayed properly
- Auto Model: Decision Tree and Random Forest are now using the latest (faster) implementations for regression problems
- Auto Model: Increased the number of rows for which local explanations are turned on by default
- Auto Model: Loading results from a folder are now adding them to the result list as well
- Auto Model: Shows the total number of feature sets and generated features on the overview as well if automatic feature engineering has been turned on
- Auto Model: The performance tab now shows the gain calculations based on the confusion matrix instead of the predicted data set
- Auto Model: New deploy button in the overview table for each model
- Auto Model: Clicking a model in the overview table will show the details for the selected model
- Auto Model: Prevent another deployment while another deployment is currently performed
- Turbo Prep: Nominal column handling is now consistent to the default behavior of Auto Model
- Turbo Prep: Sort first join keys alphabetically instead of by ID-ness
- Google Storage connection is now replaced by the more general Google Cloud Services connection that can connect to all supported Google services (Google Cloud Storage, Google BigQuery [requires In-Database Processing extension]). Just select the access scopes you want to use
Bugfixes
- Fixed all predictions being 0 in Transformed Regression when using no transformation and no z scale
- Fixed the Import Data dialog failing when trying to read an XLSX file which did not have a lowercase file ending
- Repository location chooser now opens as expected if the process is stored in a read-only repository
- Visualizations: Exporting as PDF now also works without internet access
- Visualizations: Fixed broken names in Czech Republic map
- Time Series: Added error handling if the indices attribute was also selected as a time series attribute, or as the horizon attribute
- Deployments: Fixed a bug which broke a local installation of Model Ops after the user connected to an existing remote location with email connection
- Model simulator now works with date / time columns again
Development
- Improved exception handling when Belt tables cannot be converted to example sets
