摘要
在建筑行业的投标报价过程中,如何在众多的信息中选出对最终报价影响较大的几项因素以及如何确定这些因素与最终报价之间的关系是一个棘手的问题.论文针对这两大难题提出了基于神经网络技术的解决方法.首先根据贡献变量分析理论确定出影响报价结果的9个报价因素,从而建立起基于神经网络的报价模型,然后在所确定模型的基础上改进传统的BP算法,进一步提高网络的泛化能力.从实际应用的结果可以看出,经过变量选择后所确定的报价因素是合理的,改进学习算法后的网络的泛化能力也有了很大的提高.
In the bidding procedure of construction, how to select the factors that have great affection on the bidding result and how to make sure the relationship among them are two key problems. The paper proposes a way based on the technique of neural networks to solve these problems. First, the selection theory of variables' contribution is used to identify 9 important factors that affect the final bidding result and then a bidding system is constructed based on neural networks. Second, the traditional BP algorithm is improved to advance its generalization ability. The simulation indicates that the bidding factors selected in this way are reasonable and the network with new learning error function has a much better generalization ability.
出处
《系统工程学报》
CSCD
2003年第4期348-355,共8页
Journal of Systems Engineering
基金
大连理工大学知识科学研究中心基金资助项目.
关键词
建筑行业
投标报价
神经网络
报价模型
neural networks
bidding
generalization ability
variable selection
