1. To troubleshoot above error, user can use Windows "Process Monitor" to track which DLLs are missing when loading the sklearn python library.
Below two entries are found from python processes in Process Monitor when above error occurred. "libifcoremd.dll" and "libmmd.dll" are missing in specific locations.
C:\Miniconda3\Lib\site-packages\scipy\sparse\linalg\eigen\arpack\libifcoremd.dll NAME NOT FOUND
C:\Miniconda3\Lib\site-packages\scipy\sparse\linalg\isolve\libmmd.dll NAME NOT FOUND
2. Searching through Miniconda python installation directory, the "libifcoremd.dll" and "libmmd.dll" DLL files are however found under path "C:\Miniconda3\Library\bin":
3. Copy the two DLLs ("libifcoremd.dll" and "libmmd.dll") to the respective paths shown in Process monitor, that is,
copy "libifcoremd.dll" from "C:\Miniconda3\Library\bin" folder and past it to the "C:\Miniconda3\Lib\site-packages\scipy\sparse\linalg\eigen\arpack" folder:
copy "libmmd.dll" file from "C:\Miniconda3\Library\bin" folder and past it to the "C:\Miniconda3\Lib\site-packages\scipy\sparse\linalg\isolve" folder:
4. Then re-run the Statistica python node with "from sklearn import decomposition". The "DLL failed loading" error vanished. The library is loaded successfully.