Abstract: Short-term passenger flow prediction of urban rail transit can provide important decision-making basis for relevant operation departments to adjust traffic scheduling in real time and improve operation efficiency, and provide reasonable travel suggestions for passengers. Therefore, aiming at the prediction of short-term passenger flow in and out of subway stations with nonlinear and random characteristics, this paper introduced Genetic Algorithm (GA) on the basis of Stacked Long Short Term Memory (SLSTM) network, and constructed a GA-SLSTM prediction model. The paper sorted out the historical operation data of the subway granularity of 10 minutes, analyzed the characteristics of passenger flow changes, and compared their prediction effects with GA-RNN and LSTM network models. The average value of the decision coefficient R2 of GA-SLSTM method for the predicted values of ordinary stations and transfer stations reached 0.95 and 0.90 respectively. The predicted value has a good fitting effect on the real value, and the prediction error is lower than the other two models, which proves that this method can improve the accuracy of short-term passenger flow prediction of subway.