The Development of GM (1,1) Error Toolbox Based on C Language

Abstract

In the prediction research, the main purpose is to minimize the prediction error; however, the goals cannot be fulfilled completely. Even we choose GM (1,1) model, we also need to minimize the prediction error. Hence, in this paper, we first focus on the influence parameter α in GM (1,1) model, then, analyze the characteristics of α step by step. Second, we give up the α=0.5 method, and use numerical method to find the prediction error corresponding with α value and plot the figure of the function of error. Third, for massive data testing, they show that the minimum prediction error does not occur at α=0.5, even not nearly by α=0.5. Fourth, the average prediction error for which the Class Ratio test are fail is sufficient larger than the average prediction error for which the Class Ratio test pass. Finally, after the mathematics model has been presented; we also develop a toolbox, which based on C language to assist us to implement our approach. Consequently, we conclude that the value of α is adaptive in the interval of [0,1] in GM (1,1) model.