Uncovering the hierarchical structure of text archives by using an unsupervised neural network with adaptive architecture

Abstract

Discovering the inherent structure in data has beconae one of the major cheillenges in data mining applications. It requires the development of stable and adaptive models that are capable of hsindling the typically very high-dimensional feature spaces. In this paper we present the Growing Hierarchical Self-Organizing Map (GH-SOM), a neural network model based on the self-organizing map. The main feature of this extended model is its capability of growing both in terms of map size as well as in a three-dimensional tree-structure in order to represent the hierarchicfil structure present in a data collection. This capability, combined with the stability of the self-organizing map for high-dimensional feature space representation, makes it an ideal tool for data analysis sind exploration. We demonstrate the potentistl of this method with an apphcation from the information retrieval domain, which is prototypical of the high-dimensional feature spaces frequently encountered in today's applications.