Algorithm on rule extraction based on rough set and neural network theory

Rough set method could obtain the acceptable classification pattern through removing redundant information based on no additional information given in advance. Neural network had strong self-organization and noise suppression capability. This paper proposed a method for rule extraction based on rough set and neural network combining both advantages. Firstly this paper dispersed initial data set and initiatively reduced condition attributes of decision-making table using rough set, then learned and forecasted data using neural network and filtrated noises of decision-making table through deleting unclassified data, finally extracted rules using value reduction algorithm of rough set. The experiment proves that this method is quick and effective, and can remain high robustness of neural network avoiding the difficulty to extract rules from neural network compared to traditional rule extraction algorithms.