Study And Application Of Neural Network Based On Rough Set Theory

Rough set theory has developed into one of the hot subjects in the area of intelligent information processing, which is a powerful tool to deal with incomplete and unversed information .It performs well in data mining, expert system, pattern recognition, image manipulation, machine learning, neural networks, etc. Its theory and application technologies have become an important research task in many science areas and gain much attention for its special superiorities.Actual research state, development trend and application areas of neural networks and rough set theory are summarized in this paper. Basic concepts, model and algorithms of neural network, RS and RNN (rough neural network) are also set forth in this paper .A kind of RNN model is established. It consists of rough neurons, which is the difference between RNN and typical neural networks. And its learning efficiency is higher than that of typical neural network. To the question of learning pattern complexity, the idea of knowledge simplifying is introduced to filter learning patterns .To the question of low convergence rate and defect of tending to get into local extremum existing in the learning procedure of RNN when pattern size is large and property difference of various patterns is small, genetic algorithm is adopted to optimize learning algorithm. The optimized network has rapid convergence rate and better generalization ability.In light of the combination of research production on RNN and the scientific research project of the eighth oil recovery factory in Daqing oil fields, the sedimentary facies recognition system is developed based on RNN, which realizes auto recognition of sedimentary facies.