Applying data compression technology to mobile edge computing (MEC) systems can significantly enhance the efficiency of processing a large number of tasks that require rapid response in large-scale Internet of Things (IoT) environments. In order to reduce the overall computing delay of user equipment, a multi-user task offloading strategy based on data compression is first proposed for the MEC system with non-orthogonal multiple access. Under the limitation of energy consumption, the total processing delay of multi-user equipment is minimized. For the formulated optimization problem, the deep deterministic policy gradient algorithm-based reinforcement learning method is exploited to obtain the optimal solution. The simulation results show convergence, demonstrating that the proposed transmission scheme and algorithm can find a local optimum.