Relying on the project of Xiaolangdi Dam Yellow River Diversion Project, 500 groups of shield tunneling data are selected, and shield tunneling process is taken as the sequence of data sets. Based on BP, PSO-BP and LSTM neural network model, those prediction models of shield thrust are established respectively, and three models are used to predict shield thrust. The results of prediction are compared and analyzed in detail. It can be seen from the research results: (1) All the three neural network models can well predict shield thrust, and their prediction set R2 are biggger than 0.75, the LSTM neural network has the best prediction effect which the R2 is 0.86, and the error of most prediction data is less than 20%. (2) By comparing the prediction results of those models, it can be seen that the prediction results of BP and PSO-BP neural network are generally bigger than the true value, while the prediction value of LSTM neural network are generally smaller. (3) By comparing the prediction results of those models, it can be seen that PSO algorithm effectively optimizes BP neural network, and LSTM neural network has the best prediction effect.