چكيده لاتين
Given the vast number of books available in a library, students often find it difficult to select the right book from each section. Manually choosing books can be time-consuming, making a book recommender system essential. This research aims to develop a book recommender system for the Central Library of the University of Isfahan. It is descriptive and applied in nature, focusing on examining and describing the relationship between the documents that a specific user borrows. To achieve this, all available data from the Central Library Management System of the University of Isfahan, covering the period from 1396 to the end of 1401, was extracted. This dataset contained 37823 records. After cleaning the data, the Pandas and NumPy libraries were utilized for analysis in Python. For data clustering, the Scikit-learn library was employed, using the K-means algorithm to form clusters. Normalization was then performed using the Min-Max scaling technique, and relevant clustering was conducted with k set to 2. Next, the rules and patterns of usersʹ borrowing behavior were identified through association rules. To create the book recommender system, additional data transformation and normalization were conducted using the Min-Max algorithm. Afterward, the training of the model was implemented, followed by evaluation and performance assessment of the trained model. The optimal model was selected among those that were implemented. The results indicated that users from the faculties of Literature and Humanities, Administrative Sciences and Economics, and Education and Psychology had the highest book usage. Furthermore, the findings suggested that, generally speaking, the most active users of the Central Library of the University of Isfahan are students from the Faculty of Foreign Languages. Among male and female users, female users demonstrated significantly more checkouts and had more overdue books. The most borrowed books by faculty members were also identified. Ultimately, the clustering patterns and rules derived from the data led to the development of a book recommender system that considers the characteristics of gender, member type (academic staff, students, administrative staff), and faculty. The recommender system first requests the userʹs ID and then prompts for the desired number of suggested book titles before presenting the recommendations. Based on the findings of this research, the Central Library of the University of Isfahan can review its policies regarding faculty users and prioritize the acquisition of books by focusing on the most frequently borrowed titles identified in this study.
