<![CDATA[Purpose: This study evaluates and compares the forecasting performance of traditional statistical methods (Simple Moving Average and ARIMA) and a machine learning approach (Random Forest) in predicting student applicant numbers across multiple study programs at XYZ University for 2024–2026. The objective is to identify the most accurate model to support data-driven strategic planning and enrollment management. Methods: A quantitative comparative forecasting design was applied using historical admission data from 2018–2023. Three models SMA, ARIMA, and Random Forest were implemented and assessed using Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and Mean Squared Error (MSE). Model robustness was evaluated across several study programs with different growth patterns. Result: The findings reveal an overall upward enrollment trend, particularly in Management and Informatics Engineering. Random Forest achieved the highest predictive accuracy, with MAPE values ranging from 6.17% to 17.95%, outperforming ARIMA (17.95%–33.13%) and SMA (14.5%–28.25%). The results indicate that Random Forest more effectively captures complex and non-linear enrollment dynamics. Novelty: This study provides a systematic multi-program comparison between classical time-series models and a machine learning approach within a single institutional context. It demonstrates the superior robustness of Random Forest and supports integrating machine learning–based forecasting into higher education information systems for improved strategic decision-making.]]>
