Project Estimates Based On Computational Intelligence

With the rapid development of the national economy, the old project estimation model has clearly revealed its drawbacks, and plays a certain role of obstacles and constrains of the development of the industry. The state has invested a great deal of funds for project construction, which must be paid attention to by the relevant departments and relevant personnel. The reform of the project estimation models must be accelerated to do, and it will be close to international practices. It will be an inevitable trend for the current project cost management; Project estimate is that in the investment decision-making stage, based on available information and investment estimation of indicators, experience and methods of the project to estimate the amount of investment.After the years of theoretical exploration and practical efforts, Engineering estimates become more important during the project cost management, we must ensure the timeliness and accuracy of estimates at the investment decision-making stage in order to manage the entire process of project cost in project cost place for effective control in order to improve investment efficiency and social benefits, reducing the loss of China's fixed asset investment. Although there are many methods for estimation, the accuracy of the estimation is not high. In recent years, the rapid development of neural networks and the Bp neural network algorithm which estimates the project cost has made encouraging achievements, but Bp algorithm is based on gradient descent method converges slowly, easy to fall into local minimum points of defects. The genetic algorithm is the most widely used search algorithm for optimization of one, at present, the genetic algorithm has been applied in many fields, such as function optimization, model optimization, structure optimization, industrial production, image processing. But the basic genetic algorithm is easy to fall into local optimum, in some cases too slow convergence speed, which makes the basic genetic algorithm is difficult to find the global optimum. So this paper will make the genetic algorithm and neural networks be combined so that they complement each other so as to obtain the optimal solution.The main idea of this optimization:First, randomly generate the initial neural network weights, then use the genetic algorithm optimization neural network weights, and it will optimize the obtained parameters as the initial weights of neural networks, genetic algorithm not only retains the strong global random search capabilities, but also has neural network robustness and self-learning ability, and can be both extensive mapping ability of neural networks and genetic algorithm is fast global convergence properties. The main contents:Subject the source of meaning and the theory based on computational intelligence; neural network sources, classification and characteristics of neural networks; Bp neural network algorithm and simulation in engineering; basic knowledge of the genetic algorithm and its characteristics; neural network and genetic algorithm combining the basic application has matured theory of simulation in engineering and the standard estimate and compare Bp algorithm estimates; Finally, the paper concluded.