SOM, is a neural network tool developed by Kohonen (1984,1995). They are principally used to facilitate the visualization and interpretation of high-dimensional datasets, although they may be applied to address a number of other problems in spatial analysis. In broad terms the SOM approach attempts to find a set of attribute vectors (output vectors, effectively multi-dimensional cluster centers), which are used to represent a large set input attribute vectors. The enforced spatial relationship is designed to ensure that similar input vectors are assigned to output vectors that are spatially near to one another in the output map. https://www.spatialanalysisonline.com/HTML/index.(html?introduction_to_artificial_neu.htm)