Abstract: Aiming at the real-time task characteristics of the data stream system，a dynamic priority scheduling algorithm based on fuzzy comprehensive evaluation is presented.Making use of linguistic fuzzy sets for describing uncertain factors and different priority levels，the principle of maximum membership function for determining the task priority level.Most of the task of data stream is continuous and period.There are depending relations and shared sliding windows among the tasks.The influence of these characteristics is sufficiently considered.Based on hit value ratio，differentiated guarantee ratio and CPU switching frequency，an experiment which testes the LSV performance is completed.The experimental results show that the algorithm can improve the performance compared to the classical HVF，LSF and fuzzy dynamic pre-emptive scheduling algorithms.