Numerous advances have been made in developing intelligent system. The artificial neural network (ANN) was proposed, inspired by biological neural network, to solve the problems of pattern recognition, prediction and control optimization. While ANN provides the most advanced and accurate algorithms for many AI applications, it will consume much computing time and computing power. Hence, it is necessary to design a hardware efficient ANN accelerator for the future complex computing operation. However, due to the intensive computation and communication among each ANN neuron, the interconnection between each ANN neuron become complicated as the ANN size is scaling up. On the other hand, the network-on-chip (NoC) interconnection is proven as an efficient way to solve the problems of complicated multicore interconnection. Therefore, the NoC-based ANN design is an attractive way for the ANN hardware accelerator design. To facilitate the NoC-based ANN evaluation in the system level, we present a cycle-accurate NoC-based neural network simulator, NN-Noxim, in this paper. The classification precision of the simulation output is verified by the commercial high-level neural network simulator. Consequently, the proposed simulator can be used for the NoC-based neural network design, neural network computing design, and other related researches in the future.