The Tender Offer System, Neural Network-based

Aiming at various difficulties in the bidding procedure of construction industry, the paper develops a decision-support system based on the artificial intelligence techniques such as neural networks. The following is the main content of the study:First, as known to all, the "bidding situation" that includes the factors must be considered during the bidding procedure in construction industry is always hard to make clear. The paper uses the method of layer contribution analysis to solve the variable selection problem in the bidding procedure for the first time. On the one hand, the difficulty of setting the right function during the modeling procedure is avoided, and on the other hand, the types of function in regressive modeling are extended, which makes the study of variable selection is more general. The simulation proves that not only the calculation is simplified, but also the reduced model is more coincident to the requirement of the bidding process.Second, the outputs' quality of the NN system is greatly depended on the data sent to the network as the training samples. In the real bidding procedure, the bidding situations the contractor faces are always quite different. So the generalization ability of the system is more important. In order to improve the generalization ability and to make the system more accordant to the practical requirement, the paper introduces the K-L information distance to the error function of the traditional BP algorithm. The simulation proves that the generalization ability of the system trained with the modified algorithm is much better than that of other algorithms.Based on the reasonable model, a bidding demo system is developed in the environment of Visual C++ 6.0. By the interface between the users and the computer, the demo system put the bidding system into the practical use.