Predicting factors of library traffic for UiTMCTKKT Cendekiawan Library using predictive analytics / Azzatul Husna Abdul Aziz

As organizations grow more complex, Business Intelligence (BI) and data analytics have become vital for making informed decisions and optimizing resources, including in academic libraries. The study addresses the operational inefficiencies at UiTMCTKKT Cendekiawan Library, including limited insights into student engagement, delays in material allocation, and a lack of detailed data visualization. The project aimed to predict factors of library visits, classify categories of book borrowing, and assess the level of student satisfaction. The CRISP-DM methodology was followed to apply machine learning algorithms, namely Random Forest, Decision Tree, and Naive Bayes, to the data gathered in the library which is traffic, book rentals, and questionnaires. Cross-validation was used for evaluation to ensure that the models are robust and reliable. Naïve Bayes performed best for library traffic factors with 77.20% and student satisfaction with 94.80%, while Random Forest had the best for book borrowing categories with 64.11%. Key findings highlight the role of attributes, such as study year, week 13, week 10, week 14, age, week 12, open access, and Theory Of Reasoned Action, like student attitudes, intention, and subjective norms in influencing library usage and satisfaction. To ensure the practical utility of results, the developed Power BI dashboard was evaluated by three experts using the DATUS model. This model has assessed the usability, efficiency, and operability of the dashboard, confirming its effectiveness for strategic decision-making and optimization of resources. All experts agreed on the dashboard’s ease of use, effectiveness, and efficiency, noting its clear organization and pleasant interface. They proposed a few minor modifications like improving visualizations and adjusting guidelines for better clarity. While the project succeeded to an extent in showcasing predictive analytics' potential to advance library operations, limitations were identified as static updates of dashboards and biases in the surveys themselves, which limited generalizability. Future recommendations would integrate real-time data updates, extend the period for collecting data, and refine the survey design to enhance the precision of predictions and applicability.