改进模糊神经网络的应用于投标报价_韩敏.pdf

14 5 系 统 工 程 理 论 方 法 应 用 Vol. 14 No. 5 2005 10 SYSTEMS ENGINEERING- T HEORY METHODOLOGY APPLICATIONS Oct. 2005 : 1005-2542( 2005) 05-044 -06 韩　敏, 　范迎南, 　孙燕楠 ( , 11602 ) 针对模糊神经网络规则膨胀 致的网络训练速度慢和泛化能力弱的缺陷, 提出了一种改进的基 于T -S 模型的模糊神经网络的结构和算法。网络结构包括前件和后件网络二部分, 本文在后件网络中增加 了一个隐含层以提高计算能力, 在前件网络中运用了有效模糊规则选取的方法以提高收敛速度。最后将提出 的网络结构应用于建筑工程的投标报价中, 仿真结果证明: 该网络能达到更高的误差精度、更快的训练速度 和更好的泛化能力。 : 模糊神经网络; 投标报价; 规则选取; 泛化能力 : TP 118: A An Improved Fuzzy Neural Network and Its Application in Bidding , - , - H A N M in FA N Ying nan S UN Yan nan ( School of Electronic and Information Engineering, Dalian U niv. of T echnology , Dalian 11602 , China) , Abst ract Because of the expanding of the fuzzy rules caused by the increasing of the input numbers the training speed is slow er and generalization ability of the fuzzy neural netw ork is w eaker. T his paper proposes an improved fuzzy nerual netw ork ( IFNN) based on T-S model to overcome these problems. T he : . structure of IFNN contains tw o sub netw orks premise netw ork and consequent network One hidden layer is added in consequent netw ork to improve its approximate ability and generalization ability . And in premise netw ork the fuzzy rules are chosen efficiently to speed up the train speed. At last the proposed . netw ork is used to the construction bidding system The results of the simulation indicate that it can obtain , . higher error precision training speed and generalization ability Key words: fuzzy neural network; bidding; rule chosen;