In this paper we discuss on the necessity of applying data mining operators on information systems containing a set of variables which describe the characteristics and behaviors of a specific system and could be exploited in approximating system¡¦s functionality. For the problem of function approximation, we developed a new approach combining two intelligent methods. At first we used an algorithm based on the notions of rough set theory as a preprocessor to our information system. Afterward an artificial neural network is employed as a function approximator to obtain values for decision attributes of information system while values of condition ones are passed to the network. This method has been applied to a real problem of approximating values for two hydraulic- geotechnical control variables of rubble mound breakwaters, and the results have been discussed.

Keywords: data mining, information systems, modeling, function approximation, rough sets theory, artificial neural networks

Retrieve PDF document (200807_07.pdf)

Received August 25, 2006; revised January 15, 2007; accepted February 14, 2007.
Communicated by Suh-Yin Lee.
^*Corresponding author: Gh. A. Montazer.