You are viewing the RapidMiner Studio documentation for version 9.7 - Check here for latest version
What's New in RapidMiner Studio 9.3.1?
Released: June 6th, 2019
The following describes the features, enhancements and bug fixes in RapidMiner Studio 9.3.1:
- Auto Model: Added new Quality Measures operator. This operator provides the same quality measurements used by Auto Model's Select Inputs step.
- Improved repository location parameters to correctly reference connection entries relative to a process
- Auto Model: now also optimizes the learning rate for Gradient Boosted Tree models
- Auto Model: new visualizations for hyper parameter optimization charts depending on the number of parameters
- Auto Model: shows an improved error message for the rare cases where due to preprocessing the number of different classes was reduced to 1
- Auto Model: saved results now also include a process for applying the created model on new data sets
- Forecast Validation operator now also works on nominal and date_time time series (to support Default Model Forecasts for such time series)
- Fixed a memory leak for database operators (happening in particular in parallel loops)
- Fixed an issue with stopping processes that contain loops with 10.000+ parallel iterations
- Fixed another issue that could under very rare circumstances freeze the GUI
- If a relative path in a process contains too many ../ sections to select the parent folder(s), but the top level repository folder has already been reached, additional ../ sections will simply be ignored. This fixes the problem that copying a process with relative paths from a more nested structure to a less nested structure could cause unecessary errors. Of course if the referenced data is not found, it will still result in an error at process runtime.
- Retrieve operator now shows correct warning again if the selected repository entry does not exist
- Moving Average Filter operator: Added compatibility for 9.2.1, cause the new left and right size filter parameters overwrite old symmetric filter size parameter for the simple filter type
- Fixed possible out of memory error with large Gradient Boosted Trees models
- The operator execution is now done inside a worker pool. This means that parallel Streams are executed in the same dedicated pool automatically.
- Time Series Extension: Changed tutorial keys to include operator key instead of operator name, so that the documentation project can handle the keys better