基于RS与ANFIS的投标报价决策研究

Study on Bidding Decision-making Based on Rough Sets and Adaptive-network-based Fuzzy Inference System

针对建设工程投标报价中标高金的确定这一多因素决策问题,提出了粗糙集理论(RS)与自适应模糊神经网络(ANFIS)集成的标高金决策方法.首先利用自组神经织网络(SOM)对决策系统中的连续属性值离散化后,通过粗糙集理论约简属性,去除冗余信息,计算出标高金决策系统的约简;根据这一约简,经过减法聚类获得模糊推理规则数目,确定ANFIS的初始结构,最后应用ANFIS对文章中30个典型国际工程案例的标高金进行计算和预测.研究结果表明,与一般BP神经网络相比,该方法网络规模大大减小,网络结构透明,学习速度更快,而且保持了较高的预测精度.

A new method based on the integration of rough set theory （RS） and adaptive-network-based fuzzy inference system（ANFIS） was put forward for the determination of the mark-up of the construction project. Firstly, the continuous attributes in the decision system were discretized with self-organizing map neural network. Then, reduction was found by using rough set theory to reduce attributes and eliminate superfluous data. With this reduction, subtractive cluster method was applied to generate the fuzzy inference rules, and the network structure of ANFIS was initialized. Finally,the mark-ups of 30 typical international projects were calcu- lated and predicted by ANFIS. It is concluded that the model with this method is smaller in network scale, clearer in network structure, and faster in learning speed than the normal BP neural network. Moreover, this method has high accuracy. So it is suitable to solve complicated decision-making problems, such as the biding of construction projects.