Abstract: Since the traditional multivariable grey prediction model has insufficient consideration of the coupling effect of background value and time-delay accumulative term, which leads to the low prediction accuracy of the model. Based on this, we propose a new multivariable time-delay grey prediction model with dynamic background value. The model adds dynamic background value coefficient, time-delay parameters, linear correction term, and grey action quantity term to the traditional GM(1,N) model. First, the delay periods of driver factors are determined by using the grey time-delay correlation analysis method. Second, the parameter estimation method of the model is discussed and the direct solution of the TLDBGM(1,N) time response function is given by defining the derived form of the TLDBGM(1,N) model. Finally, the model background value coefficient and time-delay parameters are identified and optimized based on the differential evolutionary algorithm. The model is applied to the problem of grain yield prediction in Henan Province. Result shows that the simulation and prediction accuracy of TLDBGM(1,N) are better than other multivariable grey prediction models. The model is theoretically more generalized. And it is shown that GM(1,1), GM(1,N), and TLGM(1,N) models are all special forms of the model for different parameter values.  

Key words:  , Grey prediction model, Dynamic background value, Time-delay accumulative term, TLDBGM(1,N), Grain yield