Mining rules from one-dimensional self-organizing maps

Artificial neural networks have been used for the mining of useful information from large data sets. Most work pursues accurate results but neglects the reasoning process. However, the practicability of many mining tasks depends not only on accuracy, reliability and tolerance but also on explanatory ability. More specifically, the discovery of patterns, rules or structures in tacit knowledge is the main objective of data mining. The self-organizing feature map (SOM) can be used within the data mining and knowledge exploratory process. However, much effort is still required to interpret the output results. This paper proposes a novel approach in which knowledge is extracted, in the form of symbolic rules, from one-dimensional self-organizing maps. The experimental results demonstrate that this proposed approach not only equips the self-organizing map with an explanatory ability based on symbolic rules, but also provides a robust generalized ability for unseen data sets.