With the development of 5G and IoT technologies, mobile edge computing (MEC) is widely introduced to perform complex computing on mobile equipment. Network latency has become a major factor limiting the performance of MEC services, and there has been a large amount of research on offloading strategies for MEC computing tasks to reduce network transmission latency. However, these studies have simply assumed that all the base stations (BS) work as edge nodes (EN) and user equipment (UE) is always directly connected to the same edge nodes throughout the edge computing process. However, in real MEC scenarios, only a portion of base stations can act as MEC edge nodes, and due to UE mobility, UEs may be under different BS coverage when using edge services. For the above scenario, a novel computing task offloading strategy is proposed in this work. In the proposed strategy, a trajectory prediction algorithm based on Social-STGCNN is used to predict the location of user equipment when the computing result return, so as to the allocation of edge nodes' and local's resources can be optimized, and ultimately reduce the overall latency of computing tasks. The simulation results show the latency in the proposed strategy is more than 20% lower than that of the existing strategies.