CAJ | 학술논문

在神经网络中引入粗糙集理论和模糊聚类方法，实现建模预测。首先用粗糙集和模糊聚类进行属性约简，去掉冗余的属性。然后根据模糊逻辑规则获取合理的网络输入层、隐含层和输出层，建立优化的粗神经网络预测模型。该模型可以有效地去除神经网络中输入层的冗余神经元，合理地确定隐含层神经元的数目，使神经网络提高收敛性能，获得更好的非线性逼近能力。仿真实验结果说明：优化的粗神经网络预测模型，可提取有用信息，简化网络结构，减少训练时间，提高预测精度。在地质样品元素的预测实验中，取得了良好的效果。
재신경망락중인입조조집이론화모호취류방법，실현건모예측。수선용조조집화모호취류진행속성약간，거도용여적속성。연후근거모호라집규칙획취합리적망락수입층、은함층화수출층，건립우화적조신경망락예측모형。해모형가이유효지거제신경망락중수입층적용여신경원，합리지학정은함층신경원적수목，사신경망락제고수렴성능，획득경호적비선성핍근능력。방진실험결과설명：우화적조신경망락예측모형，가제취유용신식，간화망락결구，감소훈련시간，제고예측정도。재지질양품원소적예측실험중，취득료량호적효과。
We incorporate rough set theory and fuzzy clustering method into neural network to realise modelling and forecasting.First we use rough sets and fuzzy clustering to conduct attribute reduction,and remove redundant attributes.Then according to the rules of fuzzy logic we obtain reasonable network input layer,hidden layer and output layer,and build the optimised rough neural network prediction model.The model can effectively remove redundant neurons in input layer of the neural network and reasonably determine the number of neurons in hidden layer,the neural network gets the improved convergence performance and better nonlinear approximation ability.Simulation experimental result demonstrates that the optimised rough neural network prediction model is able to extract useful information and simplify the network structure,it can reduce the training time and improve the prediction accuracy as well.In prediction experiment on elements of geological samples,it achieves good effects.