Recommendation Algorithm of Transformer Multi-Scene Based on Lightweight Adaptation Layer

With the wide application of personalized recommendation systems in e-commerce, social media and online news platforms, how to accurately capture the changes of users' interests in different scenes has become a key issue. This paper proposes a Transformer multi-scene recommendation algorithm based on lightweight adaptation layer, which aims to improve the efficiency and adaptability of recommendation system. By introducing time coding and parallel block design, the proposed algorithm optimizes the structure of Transformer model, so that it can deal with the time dependence in sequence data more effectively, and adjust the parameters flexibly to meet the needs of different scenes. The experimental results show that the proposed method outperforms the existing recommendation algorithms, including the traditional Transformer model and other multi-task learning models, on several public data sets. This study provides an efficient and flexible solution for multi-scene recommendation tasks.