Abstract: In the volume fracturing of unconventional oil and gas reservoirs, a large number of proppant particles are injected into underground along with the fracturing fluid, and their placement patterns will determine the propping effect and conductivity of fractures. Accurate prediction of in-fracture proppant placement patterns contributes to the optimization of fracturing design and the improvement of fracturing efficiency. At present, experimental and numerical methods are the main approaches to reproduce the proppant accumulation process and placement patterns in fractures, which are still confined by limited simulation scale, time-consuming computation and high-cost operation. In this paper, the numerical simulation results of proppant transport were adopted as data sets for input, training and testing. The characteristic parameters reflecting the process of proppant accumulation and packing were extracted, and an intelligent proxy model for the prediction of proppant placement pattern was established based on the cascade neural network. The results show that the predictions of proppant placement patterns are highly consistent with those presented by numerical simulations, while the time consumed by prediction in only 1.4‰ of that consumed by simulation. The model and approach proposed in this study have accelerated the speed of proppant transport simulation and greatly shortened the prediction time of proppant placement pattern, which will be widely applied in fracturing at field after further improvement.