The need for efficient allocation of radio bandwidth to the users in the current and future cellular networks is high. As more users are requesting for the network connections, the radio bandwidth must be fairly and efficiently allocated. The non-orthogonal multiple access (NOMA), which has been introduced for 5G networks, is useful to allow bandwidth sharing among the users, rendering a more efficient allocation and utilization of the bandwidth. However, the computational load required to determine and select the users for each of the radio resources tends to increase as the total number of users rises. Therefore, user grouping that requires lower computational complexity such as the heuristic 
methods are useful to address this problem. An ant-colony optimization has been applied in this paper to perform user grouping in 5G NOMA. The mean throughput and mean square error have been measured when testing the proposed 
scheme. The result shows that the proposed scheme produces satisfactory throughput results, close to that of the theoretical upper limit. The mean square error is also close to the lower limit and better than the existing scheme.