用神经网络进行投标报价中的变量选择

针对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这间的关系这两大难题，提出了一种基于神经网络的变量选择方法，其基本思想是：经过两次选择，消除了两个多余的变量，将最初的10个变量缩减为8个，并以这8个变量作为输入建立新的报价模型，从网络的泛化结果可以看出，经过变量选择后，网络的泛化能力有了较大的提高，表明了该方法的有效性。

A factor selection approach to bidding process using neural network

There are many factors in bidding process of construction industry, but only some of them are important to affect the final bidding results. However, it is difficult to identify these core factors. This paper proposes an approach by means of the neural network theory. After two procedures of selection, two redundant factors are eliminated. The initial ten factors are reduced to eight core factors. A new bidding model is developed with the eight factors as input, whose generalization ability is improved greatly, and this indicates the validity of the approach.