Study on parameters characteristics of NGM(1,1,k) prediction model with multiplication transformation

To further study the properties of NGM (1,1,k) model and reduce its modeling complexity, we improve this model by putting forward a formula to solve its parameters, building an algorithm for optimizing the NGM (1, 1, k) model in terms of the least modeling error, and designing a key technology for implementation of this algorithm. The optimized NGM(1,1,k) model is built accordingly. The parameter characteristics of the two models under multiple transformations and its effect of the simulation value and forecasting value are analyzed by studying the properties of multiply transformation of the two models. The research finding shows that the modeling accuracies of the two models are all in no relation to multiple transformations. The above results imply that the data level can be reduced; the process of building the two models can be simplified; but the simulative and predictive accuracy of the two models remain unchanged.