Application Research On Artificial Neural Network In The Construction Engineering Estimate Cost

The accuracy of building cost estimation impact investment decision of construction projects, and it play a major role in the cost control in the whole process of construction cycle, so it is very important to find an efficient and fast estimating method to suit the demand of the rapid expanding construction management.Artificial neural network is a adaptive dynamic system, Its function is close to the human brain, it can adapted to the environment and sum up rule, it has the very strong ability of experience Learning、association and fault tolerance, ANN can achieve more input, multiple output prediction function. These features accord with construction engineering estimate requirements. Artificial neural network has a very good application in prediction of earthquake prediction, stock trading prediction, petroleum exploration and prediction.There are many influencing factors in Construction project estimate, Carries on the analysis and induction, extraction of feature. To apply the unique characteristic of nonlinear mapping of artificial neural network to establish the relationship between features and cost, and form mapping relation, establish estimation model, realize the prediction function.The paper study deeply the theory, algorithm, characteristic feature and apply knowledge of BP, RBF, Grey BP and Grey RBF Neural Network. And build BP, RBF, Grey BP and Grey RBF Neural Network model. Study for MATLAB tools being the same with engineering, and take advantage of characteristic feature of MATLAB in what is good at calculation of large amounts of data, signal processing, program simply and program visually, program rapidly, debugging program quickly, the building cost estimation model simulation is implemented.In this study,54samples are chosen to test and analyze model, these samples are existing project between2002and2006in ShangHai, and these samples are different building types and different architectural standard. Test applicability of four models in the architectural engineering comprehensive estimation, and compare the results, RBF neural network model can make the error smallest; RBF is more suitable for construction engineering estimation.