Small Data Fusion Algorithm for Personalized Library Recommendations

Library Personalized Recommendation

The research of library personalized recommendation focuses on the personalized recommendation algorithm, personalized recommendation model, and personalized recommendation system. Ye (2021) used the LDA (Latent Dirichlet Allocation) theme model to identify the theme of book reviews and implement book recommendations based on the similarity between theme and readers. Ifada et al. (2019) used probabilistic keyword models in collaborative filtering algorithms based on book circulation records and book keywords. This, in turn, improved recommendation performance. To enhance the recommendation effect, scholars built a recommendation model. For example, Zhang et al. (2011) introduced ontology to build a personalized recommendation model that measures the similarity between books and users to achieve accurate recommendation. Yin and Zen (2019) calculated the value of readers’ interest by using borrowing time and renewal time, building a library resource recommendation model based on the change of demand time.

Based on the above research status, the library personalized recommendation research gradually pays more attention to the readers and users. However, the technical application in data analysis needs further study. The use of data has a single dimension. Data granularity and the mining of user needs are too shallow. The interaction history of users and projects contains a lot of semantic information. The existing collaborative filtering methods cannot capture hidden information. Small data has individual pertinence, seeking common ground while reserving differences and creating conditions for obtaining users’ personalized needs. This coincides with the concept of accurate knowledge service of the library. Therefore, this article attempts to achieve data fusion from the multi-dimensional user small data level, making data features interact more fully and improve recommendation accuracy.