<![CDATA[University libraries face challenges in evidence-based collection planning due to the underutilization of borrowing circulation data. This study proposes a hybrid prediction model that combines Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) to predict book borrowing patterns at the Library of Universitas Islam Negeri Sumatera Utara (UINSU). The proposed approach employs XGBoost to learn relationships among borrowing-related features and generate prediction outputs, which are subsequently used as additional input features for the LSTM model to capture temporal borrowing patterns. The dataset consists of 3,961 physical book borrowing transactions collected between 2024 and 2025. After preprocessing and feature engineering, the data were chronologically sorted and divided into 80% training data and 20% testing data using a time-based split to preserve the temporal sequence. Model performance was evaluated on the testing data using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). Experimental results demonstrate that the proposed hybrid XGBoost–LSTM model achieved an RMSE of 0.3879 and an MAE of 0.1868, indicating good predictive performance with relatively low error rates. The comparison between actual and predicted borrowing values further shows that the proposed model is capable of capturing borrowing trends effectively. The developed model can serve as the foundation for a Decision Support System (DSS) to support data-driven collection planning and acquisition strategies in academic libraries.]]>
Copyrights © 2026
