Establishment And Research Of Additional Quota For Highway Engineering Based On Back Propagation Neural Networks

With the rapid development of science and technology, and the constant upgradingof new technology, new equipment, new technologies, new materials, the existinghighway project quota is often difficult to fully meet the requirements of highwayconstruction in each region. These resulted in lack of basis for some project investmentcontrol, so that the rapid response to project valuation lost, and the project costmanagement was out of control. Therefore, the compilation of additional quota foreach region is urgently needed.This paper firstly described the research status of domestic and international quota,and analyzed the importance and practical significance of the current quotacompilation methods and the additional quota preparation; Many influencing factorswere screened to five factors which mainly effect quota by using AHP; Through theresearch on traditional compilation methods of highway additional quota, neuralnetwork quota expansion coefficient model has been established; more data whichreflect the level of production management more realistically was calculated; They laythe foundation for the rational establishment of the whole quota system.In accordance with quota compilation principle and current compilation methods,the lag of the current additional quota compilation has been analyzed, and theinfluencing factors of the quota data has been screened, so that five input neurons ofthe neural network have been identified; This paper used the structures and algorithmsof BP neural network, described the connotation of input layer quota expansioncoefficient in neural network, and simulated and forecasted quota original data byusing MATLAB; During the modeling process, the quota expansion coefficients havebeen fitted and forecasted by changing the number of neurons that implicit in themodel and training trial-and-error method of function categories; A conclusion wasfound by comparing the mean square, that the resulting prediction error is very smallwhile doing training sample inspection of BP neural network model when the hiddenlayer neuron number is 14, the training function is traingda, and the activation functionis pureline; Then the successful training of the model has been proved by thevalidation of four test samples; At last, combining with research project”the ResearchAnd Application of the Key Technology in Yunnan Highway Engineering QuotaCompilation Rules”, this paper analyzed the example by using the built model,validated the model can derive the quota expansion coefficient through three quota subheadings that Tremor Marking, Road Cutting Geocells 3D Vegetation Net, andSpecial Subgrade Waterproofing and Drainage Treatment; And the additional quota canbe finally calculated through the quota expansion coefficient. Thus the scientificity andfeasibility of this model was verified by case analysis.