Study on Rough-NN Model for Geographic Information Knowledge Discovery

This paper presents a rough neural network (rough-NN) model which is based on rough set theory and neural network technology to discover knowledge from geographic information that has high spatial autocorrelation. The main idea of this paper is to get the most concise if then rules by discernibility matrix. And a three-layer neural network to simulate the most concise rules is constructed. Inputs and outputs of the neural network are decided by the parameter-training method that is provided in this paper. This paper realizes the model with VB and presents a simulation of its use for judging drought and flood disasters in Songhua river basin. The results show that the model can quickly form the most concise rules and make right decision.