Novel weight-adaptive fusion grey prediction model based on interval sequences and its applications

With the increasing complexity of systems, it is difficult to describe a huge amount of information accurately with real numbers. To make predictions using uncertain information, this study uses interval numbers as descriptors and proposes a weight-adaptive fusion grey prediction model based on interval sequences (WAFGM(1,N) model). Unlike most interval prediction models; the proposed model uses the adaptive weight and compensation coefficient to directly predict interval grey numbers without converting interval sequences into real number sequences. Moreover, it improves the four errors of the traditional GM(1,N) model, thereby further improving model accuracy. First, by fully considering the integrity and interaction of the upper and lower boundaries of the interval, this study introduced a weight parameter to comprehensively characterize the development trend of the interval, and by optimizing the background value and adjusting the compensation coefficient, the univariate and multivariate weight-adaptive grey prediction models for interval prediction were constructed, respectively. Second, the interval development and driving coefficients, compensation coefficient, and weight parameter were estimated and optimized using the least squares method, accumulation method, and particle swarm algorithm, respectively. Finally, the WAFGM(1,N) model was constructed by combining the two models. The validity and applicability of the WAFGM(1,N) model were proved using two examples: the gross agricultural output value of Northeast China and the production of dairy products in the golden milk source areas.