Normal distribution interval grey number prediction method based on MGM (1, 2) model

Interval grey number is an important research object of grey prediction system, and the normal distribution interval grey number has become a research hotspot due to its good universality. At present, the prediction of normal distribution interval grey number was usually converted into a real number sequence by the whitenization weight function. However, it is introduced some uncertainties in the whitening process, which leads to the loss of prediction accuracy. At the same time, single variable prediction model (GM(1, 1)) was used to simulate and predict, which lacks the consideration of the correlation between the sequence development factor sequences. To solve the above problems, a normal distribution interval grey number prediction method based on Multivariable grey prediction model (MGM(1, 2)) is proposed. The distribution characteristics of normal distribution interval grey number are used to represent the sequence, and MGM(1, 2) model is used to predict and simulate. Compared with the traditional prediction model, this method can effectively reduce the mean prediction error of upper and lower boundaries from 2.26% to 1.37%, and the mean relative error is reduced from 1.79% to 0.91%.