Based On Project Cost Forecasting System Of The University Building Of Artificial Neural Network Research And Applications

The early stage of the project, in which the construction investment estimation of feasibility study is the basis of the decision-making and construction cost control for the project, often erects the most influences on the investment management of construction project. It is sorely urgent to develop an efficient and applicable method of cost estimation for construction cost in the AEC industry.The artificial neural networks (ANN) carries on the simulation to the basic characteristic of the human brain or the natural nerve network, has the very strong study, the association, fault-tolerant, auto-adapted, from the organization and the anti-jamming ability. Its powerful functions provide the potentials to estimate the cost of the architectural designs efficiently and accurately while taking several factors into account conveniently.Based on a thorough study on the conventional forecast methods of cast forecast for architectural plan, the contributing factors to the cost of architectural plan were analyzed systematically. In view of several cases from the universities in Xi'an, the engineering classification collects of architectural designs, the training regulations and the study model have been determined respectively. Finally, this paper addresses the challenge of implementing the cost forecast for the architectural designs based on several different nerve network models to eventually realize these algorithms after a compare among them.Based on the theory of artificial neural networks, a system of cost estimation for architectural designs in colleges was developed in this paper, the concrete work includes:1. Determined the configurations and contributing factors of the engineering cost Based on the conventional methods for cost estimation of architectural plans, theconfigurations and contributing factors of engineering cost were studied in this paper. Several factors affecting the engineering cost, especially the features of the engineering structure and the main sub-engineering, were figured out to serve as input variables of neural network for application.2. Studied several nerve network model and their application case of engineering cost forecast form universitiesAfter a thorough research on several models such as reverse dissemination (BP) network and radial direction primary function (RBF), the approaching function of themodel was used to simulate the variableness of the contributing factors to the engineering cost so that it can approximate the actual engineering cost more closely. Then a nondeterministic estimation method for engineering investment based on neural network was developed to provide a totally new and efficient method for investment estimation of building project in college. Simultaneously these several neural network models were compared thoroughly.3. Developed a system of cost estimation for architectural designs in colleges Based on the cost-estimating model of building project, a computer-aided system was developed. A system of cost estimation for architectural designs in colleges was analyzed and then designed following several steps: work out the client's requirements and determine the contributing factors according to different engineering structure and then build the demand model;build the neural network study model and the data analysis model to describe the system;carry out the design scheme through configuration model and flow chart;use a tool named Matlab for simulation.