The proppant placement rule in complex fractures is relatively complex. At present, most of the research is based on numerical simulation methods, which is time-consuming and laborious. In order to achieve rapid prediction of the support area, this paper uses numerical simulation methods to study the effects of injection rate, fluid viscosity, sand ratio, particle size and density and other factors on the placement effect, A neural network model for predicting the effective propping area of proppant in fractures is established. The results show that different factors have a reasonable value range, and exceeding this range will lead to poor laying effect. The variation laws of cracks at all levels with the influence factors are different, and the proppant laying of the main and secondary joints are mutually affected. After the height of the sand embankment of the main joint before branching exceeds a certain height, the higher the sand embankment height, the smaller the sand embankment area of the secondary joint,and the maximum sand embankment area of the main and secondary joints will not appear at the same time. The prediction model of effective support area in fracture based on BP neural network has an error of 3.42% ~ 6.46% between the prediction result of the model and the numerical calculation result. The model can accurately and quickly predict the area of sand dike formed after the proppant is transported in the fracture. The research in this paper is of great significance to the fracturing development of unconventional resources.