In this paper, we propose user clustering and power allocation algorithms for a precoded NOMA system. The user clustering for the strong and the weak users, and corresponding power allocation can be optimized disjointedly when SNR is sufficiently high, and this leads to a great reduction in computational complexity. Moreover this proposed power allocation algorithm can maximize sum achievable rate subject to a minimum target rate. Furthermore, from the analytical results, we suggest a reasonable minimum target rate for the weak user so that this rate can always be achieved, i.e, with outage probability 1. Using the suggested target rate, the NOMA system can avoid failing in achieving this target rate even if all power is allocated to the weak user. Simulation results corroborate theoretical results.