Coloring of the Self - Organising Maps based on class labels

The Self-Organizing Map (SOM) is a useful and strong tool for data analysis, especially for large data sets or data sets of high dimensionality. SOM visualizations map the data model dimensions to visual dimensions like color and position, thus they help exploring the SOM. Visualization can also involve the data itself so that it helps accessing information that are not available in the trained SOM, thereby enabling a deeper look inside the data. If the data comes with supervised class labels, these labels can be also involved in the visualization, thus enabling the user to have a clearer idea about the data and the structures learned by the SOM. In this work we propose a novel SOM visualization method, namely the SOM class coloring, which is based on the data class labels. This method nds a