Clustering by SOM (self-organising maps), MST (minimal spanning tree) and MCP (modified counter-propagation)

Evaluation of the cluster classification generated by a SOM is usually done by human eye. Due to the qualitative nature of this experiment, the evaluator may either overestimate or underestimate the number of clusters formed on the map. With this approach, the exact number of clusters generated by the map cannot be confirmed because of the misinterpretation of the grey-level expression. This paper presents the use of MSTs and MCP in cluster classification and reports on the application of a SOM to the chemical analysis of alloys.