Abstract: Major sudden events have emergency material demands characterized by strong uncertainty, small sample size and temporal complexity. To achieve accurate prediction and scientific scheduling of such material demand, this paper proposes an improved GM(1,1) model. This model is optimized by using Gauss-Legendre integral and GOOSE optimization algorithm, named GLG-GM(1,1). Gauss-Legendre integral is employed to reconstruct both the background value and the time response equation. Meanwhile, to further enhance the model’s adaptability, a global adjustment factor is introduced in the background value calculation. With the mean relative error as the optimization objective, the coefficient is globally optimized using the GOOSE optimization algorithm. To verify the model performance, two actual cases are selected to conduct comparative experiments with genetic algorithm GM(1,1) (GAGM(1,1)), discrete fractional�order GM(1,1) (DFOGM(1,1)), Simpson GM(1,1) (SPGM(1,1)), and GM(1,1) models. The performances of these models areobjectively evaluated by the mean relative error (MRE), mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). The results show that the prediction accuracy of the GLG-GM(1,1) is significantly better than that of the comparison models, and its adaptability in scenarios with small samples and strong fluctuations is more prominent. The research has confirmed that this model can provide reliable quantitative basis for the planning and dynamic allocation of emergency materials in Hebei Province, China, and provide effective methodological support for improving the regional disaster emergency response capability.

Key words: GM(1,1), Emergency material demand prediction, Gauss-Legendre, GOOSE optimization algorithm