Application of Data Mining in the Recommender System of Digital Libraries Based on Association Rules (Case Study: Astan Quds Razavi Digital Library)

This study aimed to analyze and examine the use of data mining techniques in the recommender system of digital libraries and information centers. By analyzing the behavioral patterns of digital library users and providing detailed suggestions, the data mining approach in this system makes it unnecessary for them to review unrelated data during the search. This not only leads to an increase in users' information requests but also significantly enhances their satisfaction with the provision of digital library services.

The current research was an analytical study of cross-sectional survey type and content analysis. This method collected the required data in four stages, and its output was analyzed using a data mining technique. Using content analysis, as in the first stage, a list of the number of transactions of user requests for books, including user IDs, manuscript titles, manuscript identification code numbers in the digital library system (manuscript database), the organization of libraries, Museums and Documents Center of Astan Quds Razavi were investigated. The data were arranged as a user column and an item (book) row. In the second step, the preprocessed raw data was further transformed into a user-item matrix, which is zero and one. In the third stage, the data output was implemented and executed using data mining technology and the implementation of association rules and FP-Growth  algorithm on RapidMiner software) and confidence (the confidence level in the desired result) were tested. In the fourth stage, the accuracy and correctness of the proposed system plan were presented.

The output of this research revealed that the association rules have a confidence level above 50% and can determine the user's access patterns, which is the best way to access the generated datasets by setting the minimum support level of 2% and the minimum confidence level of 95%, leading to 1081 new rules with conditional algorithms (if-then). If a user selects topics such as (the science of principles, Ijtihad, tradition, etc.) during the search in the digital library software, due to the history of repeated searches by previous users with the same topics by the recommender system, then titles related to the subject of principles of jurisprudence will be suggested. Also, the proof of the correctness of the proposed model showed that the first and last ones created from the new laws had thematic similarities with each other.

This study showed that various data mining techniques with the application of association rules and the implementation of the FP-Growth algorithm have high efficiency and accuracy and are suitable for analyzing the data of digital libraries and information centers to create recommender systems in order to predict user requests and make effective suggestions. One of its practical concepts is to provide a platform to significantly enhance the quality of two-way interaction between librarians and users, thereby providing optimal and beneficial services, and also to create a suitable opportunity to improve the attitude and perspective of managers in order to provide information resources that meet the real needs of users.