<![CDATA[Fluctuations in borrowing patterns within university libraries often create operational bottlenecks that diminish service quality for thousands of academic members. This study aims to enhance library resource management by developing an accurate predictive model for transaction volumes based on temporal dynamics. Utilizing the CRISP-DM framework, a Random Forest Regression model was trained on a comprehensive dataset of 52,394 transaction records spanning from 2018 to 2025. The model incorporates key independent variables, including borrowing hours, days, months, and semester examination periods. Experimental results demonstrate robust predictive performance, achieving an R2 score of 0.867, a Mean Absolute Error (MAE) of 10.89, and a Root Mean Square Error (RMSE) of 14.91. Hour of the day was identified as the most significant predictor, contributing 53.1% to the model's accuracy. These findings provide a data-driven foundation for library managers to implement dynamic staffing strategies and effective resource allocation, ensuring operational efficiency and a superior user experience even during peak demand periods.]]>
Copyrights © 2026
