It is always difficult for researchers to find a new collaborator due to the academic information overload. Although several scientific collaborator recommendation systems have been proposed, most existing methods cannot meet scholars' requirements comprehensively due to the neglect of sustainable collaborations in their schemes. Studies have shown that a sustainable collaboration has a significant positive impact on the productivity and influence for a researcher. To address this problem, we develop a sustainable collaborator recommendation (SCORE) system by utilizing the weak tie relationships brought by academic conferences for SCORE. Through defining and quantifying the conference closure, we incorporate the conference coattending relationships into collaborator recommendation system inspired by the principles of 'diversified recommendation.' The experimental results on attendees of ten academic conferences show that the SCORE outperforms state-of-the-art collaborator recommendation systems in accuracy and sustainability. Our model can be used to improve the sociability of academic conferences by recommending sustainable collaborators to conference attendees.