Short-Term Forecasting of Student Dropout Trends Using Minimal-Data Predictive Modeling

Abstract

The important and sustainable factor for the reputation and development of universities is the student retention rate. Domestic educational institutions are facing an increasing dropout rate over time, which not only affects the psychology and learning outcomes of students but also leads to unpredictable consequences for the economy and reputation of educational institutions. This study will forecast the student dropout rate using widely accepted mathematical technology. The study has transformed the initial student dropout data through a univariate forecasting method for a small data set. Through the steps in the GM (1,1) model to forecast the number of students likely to drop out for the next period. The accuracy of the model is also evaluated through the Mean Absolute Percentage Error (MAPE), the correlation coefficient (R), Gray Level Accuracy, and the posterior error ratio. The forecast results show that the trend of student dropouts will gradually increase from 2025 to 2030, with relatively good model accuracy (MAPE ≈ 2.8%, R ≈ 0.92, Gray Level Accuracy ≈ 0.97, and Posterior Error Ratio ≈ 0.4). From these results, educational institutions have a tool to forecast the dropout rate, thereby developing appropriate and proactive policies. From there, the study also highlights the role of businesses in creating conditions for students to practice and orient their careers. However, the study also has certain limitations because it has not combined many other influencing factors to make more accurate predictions.