With the exponential growth of we-media platforms, the recommendation of personalized content has become crucial for enhancing user satisfaction and engagement. However, the inherent dynamics of we-media data and the evolving nature of user preferences pose significant challenges for recommendation algorithms. This paper delves into these challenges and manages to solve them by examining past recommendation systems, such as collaborative filtering, content-based, and hybrid systems, by evaluating them with the k-means clustering model and MAP algorithm based on SQL. By examining the results, the trend in improving accuracy and stability is verified, and the characteristics of each algorithm are concluded, which may bring concerns to critics and shed light on the future development of the field.