Incorporating unsupervised learning with self-organizing map for visualizing mixed data

In previous studies, a modified SOM extended with distance hierarchies has been proposed to alleviate handling of categorical values. The model was able to take into account the semantics embedded in categorical values. However, the proposed approach required the presence of a class attribute or domain experts. In this article, we propose a model incorporating unsupervised learning of distance hierarchies so that neither class attribute nor domain experts are required in measuring similarity between categorical values. Experiments are conducted to demonstrate effectiveness of the proposed approach.