Variable Weighted Combination Forecasting Model Based on Genetic Algorithm and Artificial Neural Network

Abstract

In this paper, the variable weight combination forecasting approach which both uses genetic algorithm with global searching ability and uses neural network with nonlinear mapping ability is put forward. First, the weight coefficients are gained by means of adaptive genetic algorithm. Second, the neural network is trained by weight -obtained and the intending weighted values are predicted further. The method has character that whole weighted values is positive and the summation of weight values at same time equals to 1. At last, the variable weight combination forecasting model is built and applied into forecasting total consumption expenditure in Shanghai GDP . Simulation shows the effectiveness of the proposed approach.