The invention provides a pavement use performance prediction method, which comprises the following steps: firstly, determining a prediction index and an influence factor, sorting data, and selecting alearning sample of a neural network; adjusting model hyper-parameters, training neural network prediction models of different performance indexes respectively, and storing an optimal model; and performing performance prediction on the new data set by utilizing the established prediction model, and verifying the precision of the model. The influence of factors such as pavement structure materials,traffic axle loads, climatic environments and maintenance history is comprehensively considered, a comprehensive pavement performance prediction model is established, meanwhile, maintenance road sections and non-maintenance road sections are predicted, and great significance is achieved for performance estimation, pavement evaluation and maintenance decision making. The method is ingenious in conception, rigorous in logic and convenient to popularize and apply in the field of transportation subjects.