It is difficult for the e-commerce system to meet users individual requirements. As one of the most successful algorithms, the Collaborative Filtering Algorithm still has problems like sparsity of data and lack of authenticity in user rating. Based on the related work, a recommendation algorithm based on users actions is designed and implemented. It avoids the weakness of collaborative filtering techniques. This paper describes the recommended system’s ideas and the implementation process, and proves that the recommendation system performed well through experimtnts.