Recommendation System Based on LLM and Collaborative Filtering  
Author Wei-Shan Chang


Co-Author(s) Mingchih Chen; Ben-Chang Shia


Abstract In today's era of information overload, recommendation systems have become ubiquitous in online shopping and streaming platforms. These systems are designed to quickly match users with items or content, increasing user engagement and loyalty. Typically, recommendation systems use similarity measures to suggest items that match a user's preferences. This paper aims to optimize recommendation systems in text processing by combining them with large language models to improve performance and quality. This study consists of three main processes. First, a recommendation system is established using collaborative filtering. Second, large language models are used for downstream task training. Third, large language models with Adapter technology are employed to enhance performance and reduce training time and costs. Finally, the study evaluates the recommendation system's performance using metrics such as cosine similarity and top-5 accuracy under different algorithms. By combining recommendation system technology with large language models and using reliable evaluation metrics, this study recommends more suitable journals to researchers as an example of a journal recommendation system. The study aims to explore further optimization of large language models' applications in recommendation systems and expand their scope to cover additional domains in the future.


Keywords Journal Recommendation System, Large Language Model, Collaborative Filtering, BERT, Adapter
    Article #:  RQD28-251

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023