Research And Application Of The Grey Forecast Model

As a new transaction discipline, grey system theory which is an important branch of thescience system, is a new method which researches little data, uncertainty and poor information.Grey forecast model is one of the most important part of grey system theory. Therefore, it ismeaningful to research on grey forecast model. In the thesis, not only the classical greyprediction model GM(1,1) is researched, but also the extend model GM(2,1) and grey Verhulstmodel are also studied in depth, new improved methods are presented. Main research works areas follows:1.、The research of improving precision for the GM(1,1) forecast(1). The error of the traditional GM(1,1) model is large while the absolute value of theoriginal data’s development coefficient is big. In order to improve the simulating and forecastingprecision of the model, in the thesis, a new method which improves the background value andgrey action at the same time is proposed. And the new method is applied in the per capita energyoutput of China prediction model.(2). Because the disturbance factors often lead to that the quantitative forecast analysis andthe qualitative analysis conclusion are inconsistent in modeling process, a new kind ofstrengthening buffer operator with variable weight is put forward; the relationship between theregulation degree and variable weight of the buffer operator is studied, and the optimizationalgorithm of the variable weight is given. The experimental results show the effectiveness of thismethod.2、The research of improving the GM(2,1) forecast precisionAccording to the shortage of the traditional GM(2,1) forecast model, an improved forecastmodel by using cumulative method instead of the least square method to reduce the modelmorbidityz (1)instead ofx (1)and the gradient descent method instead of the initial value ispresented. The improved model is applied in practice; the result shows a higher precision andstability. 3、The research of improving precision for the Verhulst forecast model(1). According to the problem of derivatives albinism, an improved grey model is presented.Firstly, a new definition of Verhulst model with variable parameters using Lagrange mean valuetheorem to calculate the value of grey albino derivative is got; secondly, the genetic algorithmwas used to solve parameters. Example analysis show the new Verhulst model is of higherprecision.(2). Because of the defect that the traditional non-equidistance grey Verhulst model severthe link between difference equation and differential equation in grey modeling mechanism, animproved non-equidistance grey Verhulst model with the definition of variable parameters basedon the genetic algorithm is proposed in order to further improve the precision of the model.