Research On Construction Cost Estimation Method Based On Mathematical Models

Engineering field of project cost contains a lot of content, in order to create more value, practitioners must improve work efficiency. Therefore, how to estimate the cost of construction quickly has become the important research contents. In the stage of project feasibility study, we need to estimate the cost of project as the basis price for bidding in the future. Thus, in order to control cost of investment, we must estimate the cost of project in the early stage of engineering construction. How to establish the model of evaluation and prediction accurately has become the real problem that we need to solve, there are many methods to estimate the project cost, but these methods have their defect, the accuracy is not high. How to ensure the accuracy of estimation while retaining the speed of estimation has become an important thing that we need to seriously study.There are many kinds of traditional valuation methods, but the accuracy of calculation results is not high, and the error is large. With the continuous development of computers, in order to solve engineering problems, people will combine mathematical models to computer gradually to find a satisfactory valuation models. This paper mainly describes the fuzzy math, gray theory, neural networks, fuzzy mathematics and gray theory can solve the problem of similarity effectively, and play an important role for project cost estimation, the neural network training and testing of samples has promoted rapid evaluation of construction projects. Three methods are used in project cost, the error can be controlled within the allowable range. Through a large number of engineering example, this paper establish a database, combines theory to practice, picks out the main affecting factors of the project cost, application of fuzzy mathematics and gray theory were used to estimate the value of the project cost. On this basis, this paper put forward the gray fuzzy neural network model, according to fuzzy degree of nearness and the grey relational degree, find out the most similar engineering project as BP neural network training samples, and the sample data are quantified and normalized. Through many times of calculation, this paper constantly determine weight and threshold value of the network until satisfied. According to results of calculation error, we learn that the accuracy of the gray fuzzy neural network model, which overcome the disadvantages of previous methods, is higher than the separate use of the fuzzy and gray theory model. The gray fuzzy neural network model are more flexibility and more satisfying, which achieve the objective of the rapid project cost estimation, and provides a new solution for research on the cost of construction.The correlation function of BP artificial neural network are set in MATLAB neural network toolbox. This paper introduces functions of NN Tool in the neural network toolbox through engineering examples, describes how to use NNTool on construction projects for training and testing, and each parameter represents the significance and the role of the sample in the process of training, the operation steps of the neural network toolbox steps is described in detail.