Improvement To Gray GM(1,1) Model And Its Application

Grey system theory as a new intersect discipline, is a new method of minority? poor information and uncertainty question, it is an important branch of system science. The grey GM (1,1) model is an important part of grey system theory; it is also one of the main research content. This dissertation makes an intensive study of the gray GM (1,1) model, and we has made some innovation on the basis of predecessors' work, even we use it to solve some practical problems.In this paper, we research for one of the most commonly used model, which is grey GM (1,1) model. We found that the defects existing in the traditional grey GM (1,1) model. The initial value to select the initial value of actual sequence is not reasonable when it formed prediction formula. The reason is fitting curve doesn't necessarily used by the first data point. The first value of the actual sequence is one of the most original data. Then this data is not close to the future. What's more, it isn't getting by accumulation generation, and it hasn't strong regularity. So it's necessary for us to discard the traditional idea to select the initial value of actual sequence. We allow to choosing other data, thus we can get the higher prediction formula. As the development coefficient value of a, when the absolute value of a is larger, the deviation of model is bigger, and it can't be used for medium and long-term forecasting. So we should improve the formula background value, then the optimized GM (1,1) model can suitable for various development coefficient, especially when the absolute value of a is larger, it can also be used for medium and long-term forecasting,and we can get a high precision.In this paper, we put the initial value and the background value together. And effectively prevent model error by these two factors, so we get a high precision of the GM (1,1) model. This new method was applied to our national TV production forecast model. We get a little increase on the basis of predecessors' work. And this method has a certain practical application value.