Knowledge Studio 10.8.3 adds more integration points with Knowledge Hub, Outlier Detection Improvements, and new Variable Transformation features.
Improved Integration with Altair Knowledge Hub. All types of predictive models are now supported
Altair strengthens the integration between predictive analytics and data preparation by publishing more models created in Knowledge Studio into the Knowledge Hub. Included are: Strategy Trees, Scorecards including the Scorecard editor, Deep Learning (Neural Networks), more regression and regularization models.
Outlier Detection Improvements
The Univariate Distance method replaces the Standard Deviation method for outlier detection. Aside from selecting variables to be tested for outliers, users can choose from multiple location and scale measures. Users can also define grouping variables to calculate the measures and detect outliers within groups rather than the entire dataset.
Multiple Variable Transformation Features added
Includes: Standardization helpers to rescale numeric variables, either one or multiple variables at a time; Safe Division to better handle integer arguments and extreme cases such as division by zero and the infinity value; Infinity symbol and support for using the infinity function in SQL code to designate positive or negative infinity.