Viscovery SOMine 8.1 Release

Viscovery SOMine 8.1 is a minor feature release. These are the notable changes.

The Group Profile window can now show the adjusted P-value in a separate column if multiple testing correction is enabled.
The formula language now has a case statement and an iferror() function.
The decimal point and field separator can now be selected for text files and CSV files in the Import Data step.
The number of lines to skip at the top of text, CSV, and XLSX files (not XLS files) can be selected in the Import Data step.
The Performance Measures in the Check Application step of the Apply Classifier workflow now show both an average per case and an average per class of accuracy, precision, sensitivity, and specificity ratios.

Title bars and color scales of map pictures scale with the picture size.
Cluster “separators” are now called “Cluster boundaries”.
Lengthy computations in the Statistics window no longer block interaction with the software.
In some edge cases, Connectivity clustering cannot produce small cluster counts. Now these are no longer offered in the indicator chart.
In new projects, SOM training now uses the traditional algorithm by setting the training temperature initially to 0.
In new SOMs, the color scales of attribute pictures are now initially extended to “nice” round values.
When the user specifies bounds for the color scales that are inside the numerical range of attributes, the presence of values outside the visible range is now indicated on the color scales.
A double-click on a defined nominal value in the Define Nominal Values for <attribute> dialog no longer deletes the value.

When priority groups are defined, the resulting scaling is updated when the data changes. (But this happens for old projects only after the Prioritize Attributes dialog is entered.)
Fixed incorrectly labeled vertical axis of the scatter plot when a nominal attribute was selected.
Fixed that Apply Model steps operating in parallel would sometimes be aborted with a “duplicate name” error.
Fixed that application data marts used in Apply Classifier and Apply Predictor workflows can grow so large that they become inaccessible without module Enterprise Data.
Fixed that occurrences of duplicated nominal values in a multi-valued nominal attribute were counted more than once per data record. This affected only the counts and percentages in the Define Nominal values for <attribute> dialog.

The project file format was updated. Projects written by this version cannot be imported in earlier versions of Viscovery SOMine.

Viscovery SOMine 8.1.1 is an important service release because two serious data loss problems have been fixed.

The project file (*.visdom) could become unreadable when a Logistic regression with R was selected, Regularized was enabled, but then a Linear regression was computed. This has been corrected.
When data marts were present that had repeated parts separated by an underscore in the name, then they could be destroyed when a data mart was created with a name that is just that part.
The partition selection was not heeded by the scatter plot tab in the Statistics window of a workflow step.
Connectivity cluster indicators were not stored in the SOM file if the clustering name contained Unicode characters outside the Basic Multilingual Plane.