hSOM: Visualizing Self-Organizing Maps to Accomodate Categorical Data

Kohonen’s self-organizing map is an unsupervised machine learning method designed to preserve the topology of its input space. Although this method has been used to efficiently summarize multidimensional data, the visualization of its constituent data has received less attention. We propose a method of addressing the visualization problem by augmenting a classical self-organizing map visualization to include an embedded histogram and evaluate its utility in depicting the self-organizing maps’s constituents categorized by a discrete variable.