A novel multivariable grey prediction model with different accumulation orders and performance comparison

Abstract

The uniformity of the accumulation orders of the dependant and independent variables in traditional multivariable grey prediction models ignores the differences in physical properties amongst the variables, which results in unstable or even abnormal model performance. To this end, a new multivariable grey prediction model with dual orders is proposed by differentially defining and optimizing the accumulation orders of the dependant and independent variables. The new model expands the functional structure of the traditional multivariable grey prediction model, improves the preprocessing effect of the accumulation orders based on the original sequences, and enhances the modelling performance of the multivariable grey prediction model. A comparative analysis of three cases verifies that the new model outperforms the other two mainstream multivariable grey prediction models, which confirms that the dual orders of the novel model are reasonable and effective.