Weight Optimization Of FSE Rating Modelbased On Neural Network

Fuzzy comprehensive evaluation method is a kind of comprehensive evaluation method of macro fuzzy mathematical reasoning,qualitative evaluation and quantitative evaluation of some things with ambiguity based on the membership degree theory of fuzzy mathematics.It can better handle some of the fuzzy,difficult to quantify the affairs.The key of this method lies in the determination of the weight vector.At present,Most of the methods in the process of weight determination are difficult to avoid the impact of human subjective factors or the existence of high computational complexity of the method,and it is easy to appear inconsistent with the actual results.The neural network method to determine the weight is the use of neural network learning or training to acquire knowledge,and stored in the connection weights between neurons,then using the correlation coefficient calculation formula of the relevant information to reproduce,ponder,refine evaluation index of the objective law,so as to get the weight of related evaluation indicators.At present,BP neural network method to determine the weight of the method has been widely studied,but few people study the use of RBF neural network method to determine the weight of the method.Compared with the BP neural network,RBF neural network has the advantages of fast convergence speed,not easy to fall into local minimum.In this paper,RBF neural network method is used to determine the weights by the construction of the network structure and specific network parameters optimization design,so it can play a good effect on the determination of weight problem.Firstly,a fuzzy comprehensive evaluation model(FSE)based on neural network is constructed.The quantitative data of the rated transaction is used as the input of the neural network,and the final rating result is taken as the output of the network.The RBF neural network three-layer structure is effectively combined with the fuzzy comprehensive evaluation method flow step by means of neural network training.In order to verify the validity of the rating model,the fuzzy comprehensive evaluation model based on BP neural network and the FSE rating model of this paper are used to evaluate the fuzzy comprehensive strength respectively.The FSE rating model based on RBF neural network is proved to be more effective.Secondly,this paper puts forward a method to optimize the weights by using RBF neural network,which is based on the optimization design of network parameters,and constructs a relatively stable RBF neural network.The results of the sample data training are calculated by the inverse matrix formula of the correlation coefficient,it gives full play to the advantages of RBF neural network training speed to ensure objectivity,determine the index weight,improve the efficiency of weight determination.It is of great significance to optimize the weight of fuzzy comprehensive evaluation.Finally,this paper compares the optimal weight of the rating index with the grading index weight determined by the analytic hierarchy process(AHP),and proves the feasibility and effectiveness of the method based on RBF neural network to optimize the weight method by comparing the simulation results with the final enterprise fuzzy comprehensive rating.It can effectively overcome the analytic hierarchy process in the subjective weight and computational complexity of the algorithm defects.