Analysis of undergraduate employment trends using an enhanced grey prediction model

To overcome the drawbacks of the conventional GM (1,1) model in forecasting the employment rate of university graduates, this study advances a revised GM (1,1) grey prediction model. By combining dynamic weight parameter optimization and high-order polynomial residual correction techniques, the employment rate time series data are modeled and predicted. By introducing the λ weight parameter (optimal λ=0.01) to optimize the background value construction and combining it with the high-order polynomial to fit the residual sequence, greatly boosting the model’s forecast precision. Through the analysis of the employment rate data of college students from 1997 to 2021 (supplementary data from 1997 to 2014 is added to expand the time span— covering multiple economic cycles such as the 2008 global financial crisis and the 2015 supply-side structural reform—thereby improving the stability of model training and avoiding over-fitting caused by a narrow data range), the improved model has demonstrated outstanding performance in employment prediction. In comparison with the conventional GM (1,1) model, RMISE decreased by 3.2% (0.1812-0.1754), and MAPE decreased by 7.1% (0.14% - 0.13%). Predictions indicate that the employment rate of undergraduate graduates will show a slow downward trend from 2022 to 2025, which holds significant reference value for the planning and decision-making of employment work in colleges and universities.