Prediction of the Scale of Graduate Education in China Using a Combined Weighting GM(1,1) Model

Accurately predicting the scale of graduate education in China holds practical significance for formulating reasonable educational policies. The GM(1,1) model performs better in forecasting uncertain systems with "small data and limited information." However, the model has drawbacks, such as excessive reliance on the smoothness of data sequences, background values, and parameter estimation methods, which can lead to significant prediction errors. Therefore, in this study, we first improve the smoothness of the data sequence using a function transformation method, and then apply the traditional GM(1,1) model, the unbiased GM(1,1) model, the background value-optimized GM(1,1) model, and the combined weighting GM(1,1) model to predict the scale of graduate education in China. The prediction performance of these models is compared, and the best-performing combined weighting model is selected to forecast the scale of graduate education in China from 2023 to 2026.