Classification using topologically preserving spherical self-organizing maps

A new classification method is proposed with which a multidimensional data set was visualized. The phase distance on the spherical surface for the labeled data was computed and a dendrogram constructed using this distance. Then, the data can be easily classified. To this end, the color-coded clusters on the spherical surface were represented based on the distance between each node and the labels on the sphere. Thus, each cluster can have a separate color. This method can be applied to a variety of data. As a first-example, we considered the iris benchmark data set. A boundary between the clusters was clearly visualizible with this coloring method. As a second example, the velocity (first derivative) mode of a Plethysmogram pulse-wave data set was analyzed using the distance measure on the spherical surface.