The Research Of Genetic Neural Networks In Bidding

Since there are many problems arise from the process of bidding in construction industry, a system which can provide the strategy for bidding is elaborately designed and explored in this thesis through the combination of the usage of Genetic Algorithms and neural network.The following are the whole process of research and main findings:Combine Genetic Algorithms and neural network to produce a model to overcome the difficulty of judging the bidding situation during the process of bidding in the construction industry. In this may, make up the imperfection of BP, which is low efficiency and small localization .The findings after the experiment show this model of bidding practically meet the need of bidding better.The Genetic Algorithms is good at overall searching while BP has small localization. The usage of Genetic Algorithms can refine the imperfection in the process of study on net. The following is the steps of the experiment: "m" groups of weight factor are selected randomly and are distributed into "m" nerve net . Then deliver the samples of "0-1" Digitalization processing training sets samples to net. Then get "m" groups of figures. Use real-number coding and take the Neural network weight factor as the chromosome of GA. Produce a gene group, use Genetic Algorithms to choose the best one. Make the parameter combination close to the best parameter combination. At last use BP to adjust them in detail. So that combine the Genetic Algorithms and neural network to complete the production of the system model of high-speed bidding .Based on the reasonable model-building, use software "MATLAB" to get the achievement at the beginning. And take a lot of figures from project experiment as the sample to practice on the net, at the same time lots of experiments have been done to test them.