The Construction of Index System Based on Improved Genetic Algorithm and Neural Network

Abstract

Artificial neural network (ANN) and genetic algorithm (GA) have both prevalent uses in large area. Along with the development of technology a method based on the combination of Artificial neural network (ANN) and genetic algorithm (GA) aroused. Now there is not a quantitative way on the problem of constructing the index system. In such a case, the paper uses the combination of Artificial neural network(ANN) and genetic algorithm (GA) to solve this problem. This paper firstly establishing feedforward neural network model and make sure about the input and output variables. Secondly improved genetic algorithm is used to solve the problem of network weight and threshold value which is constitute by three steps real codes, random selection and Genetic Manipulation of Chromosome. Moreover as it know to all, error back propagation(BP) algorithm is effective in local searching so adding error back propagation(BP) algorithm to genetic algorithm is a good way to get the satisfying result. Thirdly the paper gets the output of index effectiveness. Thirdly according to the entropy theory that the summation of effective value which could be involved in the index system should be larger than a certain critical value, the paper screened out the final index. Thus, in theory, gives a quantitative method of constructing the index system.