With regard to the limited computing resource and the constraints of time delay of mobile edge computing (MEC) server, how to offload the complex computing tasks to mobile edge computing servers reasonably to carry on the data storage and calculation processing so as to both shorten completion time and reduce terminal energy consumption is the significant research contents of resource offloading algorithm. In this paper, a resource offloading algorithm aimed at the single cell multi-user scenario in edge computing with deep learning theory is proposed, considering the case that computing resources can be divided arbitrarily to decide whether to offload or not, in which the system models are established respectively taking delay and energy consumption as optimization objectives. The simulation results show that the resource offloading delay and energy consumption of this algorithm are significantly improved compared with the traditional resource offloading algorithm.