<img src="GisSOM-logo.png" height="200">

GisSOM

GisSOM performs unsupervised self-organizing maps (SOM) clustering to a given dataset. Results are presented in the SOM coordinates, and also spatial coordinates if given. Also scatter and boxplots are used to visualize the per cluster data distribution.

Installation

To download the installer for GisSOM, go to the Releases-section (https://github.com/gtkfi/GisSOM/releases), and download the GisSOM_Release.zip -file, which contains the installer along with the user manual and license information.

If you wish to use an upgraded version of GisSOM for non-commercial purposes, please contact johanna.pesonen@gtk.fi. The current version (1.6) is not publicly available, but it is significantly more stable and fluent than the version (1.3) available on this GitHub site. New features have also been added. The upgrade was done by the Geological Survey of Finland in the EIT Raw Materials DroneSOM project.

Features

• Spatial or non-spatial input data
  • CSV and GeoTIFF Results in SOM and geographic coordinates
  • CSV, txt and image files
• Interactive plot
  • highlighting data associated with a specific cluster or SOM cell
  • relation between SOM and geographic space
• Easy installation
  • Built-in Python scripts

<img src="/Screenshots/som_2.png" width="800" height="400">

^{Som results plotted in somspace}

Screenshots

Data preparation:
<img src="/Screenshots/som_1.png" width="800" height="400">

Results in Geospace:
<img src="/Screenshots/som_3.png" width="800" height="400">

Interactive plot:
<img src="/Screenshots/Interactive_plot.gif" width="800" height="400">

About the code

GisSOM is implemented in C# and XAML using the Windows Presentation Foundation (WPF) framework, according to the Model-View-ViewModel (MVVM) design model. Computational tasks related to data preprocessing, som calculation and visualization of the results are handled by Python scripts. The software uses Python version 3.7, and the Python scripts and the python environment are bundled into executables using PyInstaller.

Development

If you want to develop GisSOM, or use it or parts of it in your own projects in accordance with it's license, you are welcome to do so. Just clone or download the code, open the solution with VS 2019, set up the Nuget packages, and you should be ready to go. If you want to edit and run your own versions of the python scripts, you're going to need a working installation of Python 3.7, with all the required libraries installed (required libraries are listed in the Tech Spec file, under the GisSOM/SomUI/Documents folder). GisSOM uses bundled python executables by default, but this can be switched off by setting the usePyExes-variable in the SomTools class (rows 44 and 45 in version 1.1.0), and providing your Python installation path in the pythonPath-variable.

private string pythonPath ="C:/Your/Python/Path/pythonw.exe";
private bool usePyExes = false;

^{All set for debugging!}

Contribute

If you want to contribute to GisSOM, you can just fork the repository, and submit a pull request with your changes.

History

21.12.2022 GisSOM version 1.3.0
NOTE: Running the software requires read/write permissions, so make sure the installation path is to a folder with these permissions.
-Rework data preparation stage
-Add more informative errors on bad input data

15.03.2022 GisSOM version 1.2.4 release
NOTE: There's a known issue when using .TIF input,the software will give warnings on proj.db and PROJ installation. The software can still be used despite these warnings.
-Added Cancel button for som calculation
-Upgraded python version to 3.9 and updated used python packages
-Fixed a bug with heades of hexagonal som output plots
-Added geospace output data files for nonspatial data (with dummy coordinates)
-'nan' is now a viable data header
-Fixed bugs with CSV-scatter dataset plotting

04.01.2022 GisSOM version 1.2.3 release
29.11.2021 GisSOM version 1.2.2 release
30.09.2021 GisSOM version 1.2.1 release
23.09.2021 GisSOM version 1.2 release