Research Based On Genetic Algorithm Optimization On The Application Of BP Neural Networks In Tender Offer For Construction Projects

With the rapid development of China’s construction industry, more and more domestic construction enterprises joined into the fray of international project contracting, construction engineering bidding market competition intensified. To develop a reasonable tender offer has a pivotal position in the project tender.The bid price of construction projects is a process fraught with uncertainty. The uncertainty of the tender offer is subject to the project size, complexity and technical content, and many other factors, but also by time, resources, and environmental factors. How to predict the bid price basis of the information resources effectively becomes an important and pressing problem. It has a strong theoretical and practical significance.Companies at this stage use offer decision-making model based on probability theory, game theory-based price model and Analytic Hierarchy Process (AHP). The main data indicators involve many uncertain factors affecting the tender offer, with the relationship complex. The prediction of the tender offer is actually a nonlinear problem. BP neural network is a robust, non-parametric model to solve the problem of multi-factor. It has a strong nonlinear mapping ability, learning ability, high classification accuracy, and is able to handle the untrained interference or incomplete data, showing good fault tolerance, adaptability, dynamic performance and generalization ability. But it also has some limitations. Since the initial weights and thresholds of BP neural network is random, the network may appear to a local minimum or not convergence. To compensate for its shortcomings, this paper proposes to use genetic algorithm to improve the BP network. It uses the initial weights and thresholds through population evolution of the global search iteration by genetic algorithm into the BP network, as the initial weights and thresholds of the network, making the network accuracy and training efficiency greatly improved.The main content of this paper include the following:1. Introduce the construction works related to tender offer theory. The article describes the whole process of the tender offer, including the offer decision-making and strategy skills which applied to the current tender offer market, so it is convenient to compare with the artificial neural network modeling approach. And it also establishes the index system of the bid price factors. 2. Introduce the basic concepts of artificial neural networks, the development process and its characteristics, and several popular neural network models. And make a detailed analysis of the BP neural network for the feasibility of application in the tender offer.3. For the lack of BP network model, the article proposes the use of genetic algorithms to optimize the BP network initial weights and thresholds so as to achieve the overall requirements, making a more accurate price model. Then, it introduces the principle and steps of genetic algorithm.4. Build decision-making model based on genetic algorithm optimization BP neural network in the tender offer. Collect30sets of tender offer data from a consulting firm in Yangzhou sorting, normalization and then using MATLAB software to train the network to establish a BP neural network which can predict tender offer. Compared to the BP network optimization based on genetic algorithm and a simple BP network improves accuracy and convergence rate of the former than the latter one.