Abstract: Accurately forecasting the number of enrollments in higher education in China can optimize the allocation of educational resources and ensure the sustainable development of the education system. Unlike existing grey prediction modeling approaches, this paper proposes a novel concept, i.e., the accumulated generating sequences preserve morphological features similar to the original data while enhancing sequence smoothness, and build a novel New-Information-Priority Grey Bernoulli Model (NIPNGBM) for predicting the number of students enrolled in higher education with different trends. Then, we use the number of students enrolled in higher education per 100,000 population from 2010 to 2021 in Beijing, Guangxi, Xizang, and Ningxia as examples, testing the adaptability of the new model to various sequence patterns, such as L-shaped, rising, inverted U-shaped, and S-shaped sequences. And the last block validation method in sample segmentation is used for hyper-parameters optimization. Results indicate that the proposed model demonstrates higher prediction accuracy and lower prediction volatility, compared against LSTM, Holt-Winters, ARIMA, LSSVR, and NGBM models. Furtherly, the new model is applied to forecast and the development trends of the number of students enrolled in higher education across 31 provinces in mainland China in the next seven years are analyzed from different time points, population thresholds, and different regions. This concept that grey prediction models balance both the shape and smoothness of the original sequence can be extended to the grey predictive modeling system, making the cumulative generation and the construction of new model more compatible, thereby expanding the application scope of the model.

Key words: Data trends and patterns, New-information priority, NIPNGBM, Model’s properties, The number of students enrolled in higher education