Abstract: With the development of science itself, science of science has become research in recent years. The scientific cooperation network which is an important structural foundation of scientific research organizations and knowledge dissemination has attracted wide attention from scholars. Under this circumstance, the formation of cooperation and the weight of cooperation in the scientific cooperation network have become very meaningful research issues. This paper proposes a link prediction and weight prediction methods based on multiple motif features and machine learning framework, and compares the experimental results with several classical methods. It is found that the proposed methods can effectively improve the accuracy prediction: up to 8.9% in the link prediction and 59.6% in the weight prediction. This paper helps to predict the possibility of scientist collaboration in the scientific research network and their cooperation weight, and then to explore the profound impact of the structural characteristics of the scientific cooperation network on the scientific research output and teamwork of scholars.