Research of GNNM(1, N) Based on Self-correlation Theory and Its Application

In this paper, based on self-correlation theory and GNNM(1,N), the new forecasting approach is put forward. First, original data sequence is analyzed by means of self-correlation theory and is divided into N data sequences according to the prominence of self-correlative coefficient. Second, the generated data sequences are modeled by means of GNNM(1,N). The quantitative relations among the model parameters and the forward neural networks' weights are given. Third, the learning algorithm of the grey GM(1,N) neural network is presented. The GNNM(1,N) can improve GM(1,N) model's precision because learning error of the GNNM(1,N) can be effectively controlled. At last, the method is used to build model of total residence number in Shanghai city, P.R.C. The results of the example show that the model has by far higher modeling and forecasting precision.