Aiming at the high time complexity problem of the offloading strategy in 5 G mobile edge computing（ MEC） scenarios

a computing offloading algorithm HCOGA combined with deep learning is proposed. This algorithm models mobile services as a workflow model

and jointly optimizes the overhead of the workflow model through improving non-dominated sorting genetic algorithms（ NSGA-Ⅱ） . In order to adapt to the 5 G real-time environment

HCOGA algorithm fits non-linear relationship between input and output of NSGA-Ⅱ through deep neural network to form a complete offloading prediction model

which can make offloading decisions quickly and accurately. Simulation results show that the HCOGA algorithm has a shorter running time and can effectively reduce system overhead.