We propose an indium gallium zinc oxide (IGZO) nanofiber based photoelectric synapse. Long-term potentiation and depression emulations are realized by exploiting optical and electrical stimulus as the excitatory and inhibitory inputs, respectively. Significantly, IGZO nanofiber-based photoelectric synapse exhibit multilevel characteristics (up to 10 bits) with low updating energy (~1.0 fJ). Furthermore, an artificial neural network (ANN) based on IGZO nanofiber photoelectric synapse is built and evaluated through simulations. The performance indicates more than 93% accuracy in recognizing the standard MNIST handwritten digits, showing the great potential for high-precision neuromorphic computing by the IGZO nanofiber photoelectric synapse.