Knowledge discovery applications can be roughly divided into two categories. To begin with, flexible environments which have a multitude of methods and good database tools to handle the data with. The user of such application is a data mining specialist who has a good grasp of the possibilities and limitations of the different kinds of mining methods. Secondly there are specialized applications which are intended for end-users of the results. They perform specific procedures on the data and the user is not required to be intimate with the mining methods involved. Typically applications of the second kind are based on experiences and experiments on applications of the first kind .
As odd as it may seem, the SOM has not been widely used in data mining. This may partly be due to the fact that the KDD community is heavily leaning towards artificial intelligence and not on neural networks. The data mining tasks usually aim at rules extraction, to which the SOM is not directly applicable. Another reason is that the SOM is a relatively new technique. Although recognized as very important by the neural networks research community, it has not yet gained the kind of outstanding status as the multi-layered perceptron (MLP), the most widely known neural network technique.
The application that was implemented as part of this work, ENTIRE, belongs to the first category by being a general-purpose data mining tool. However, being in its first stages of development it concentrates solely on the use of the Self-Organizing Map. The current version of ENTIRE is a prototype program, meant to be used as a platform for the use of the SOM in data mining. It concentrates on providing the basic properties and visualization tools of the SOM, along with some essential data pre- and postprocessing abilities. ENTIRE was built on top of the SOM_PAK program package . The GUI was built using the XForms graphical user interface library for X windowing environment .