The output neurons of SOM with 101 sample vectors. (A) The U-matrix with a shading intensity of the level of activation; (B) Partitioning of the 50 hexagons into 4 clusters by the K-means algorithm; (C) SOM classification based on the similarity of CPUE of fish species/species groups; and (D) Hierarchical clustering (Euclidean distance on the vertical axis), based on the similarity between SOM neurons.