Sponsoring companies
SOM-method
Self-organizing map (SOM) is used as a data mining method to integrate different data sets and create clusters of similar areas. SOM is an unsupervised artificial neural network that projects n-dimensional data to 2-dimensional SOM lattice. The usability of SOM comes from its topology preserving nature: the data objects that are clustered in the original n-dimensional space will be projected to SOM cells that are close together. As SOM preserves topology, it can be used for clustering. This can be done by visually studying the U-matrix, which is the difference between one SOM cell and its neighbours, or applying another clustering method to the SOM cells. The clusters can be described by visualizing the original dimensions on the SOM lattice.
The clusters resulting from SOM utilizing both 2D and 3D data and their connections to geology are studied in the project to discover geological knowledge of the area. In addition, the clusters are combined with known deposits in the area for prospectivity mapping.
Clusters visualized in SOM space and geographical space
