Personalized Library Book Recommendations Using K-Means Clustering and Association Rules

Abstract

The library serves as the primary information source and has a significant impact on raising educational standards. This study aims to improve library service quality by developing a personalized book recommendation system. The personalization of the recommendation system is the design of a prediction model for books that each user will borrow based on the user's interests, behavior and other related information. The novelty of this study lies in the combination of K-Means Clustering and Association Rule techniques to create a more accurate recommendation model tailored to each user's preferences, behavior and other relevant information. One common issue in building recommendation systems is the Cold Start Problem, which refers to the challenge of making accurate recommendations for new users or items with little to no historical data. Therefore, K-Means Clustering is utilized to segment users based on their borrowing patterns, which helps address the Cold Start Problem by recommending popular books within each cluster. User information from book loan transactions is used as input for the clustering model. Next, a recommendation model for each cluster will be made using the Apriori algorithm. Apriori was chosen for its simplicity, low computational cost and ability to efficiently identify frequent patterns, making it ideal for large datasets. This study trials the number of clusters between k=3, 4, 5 and 6. The best results for the recommendation model used a combination of 6 clusters and the Apriori algorithm. The application of the clustering method can improve the recommendation model with a difference of 3.61% average accuracy, 0.67% average precision, 3.61% average recall and 0.49% average F1 score compared to the recommendation model without clustering method.