With the increasing demand for engineering construction in the seasonal frozen area and the background of the Belt and Road Initiative, the frozen soil constitutive model should be studied in depth. At present, the constitutive prediction model of frozen silty clay has many problems, such as complex formula, single model application and poor prediction ability. Random forest optimal model hyperparameter input was very difficult. Particle Swarm Optimization (PSO) was used to optimize the parameters of the number of neurons, dropout and batch_size in the Long-term and Short-Term Memory network (LSTM) structure. The optimization results were 61, 0.09 and 95 respectively. The results showed that the strength tended to be stable after 6,9,6,9 and 9 freeze-thaw cycles under initial moisture content = 25, 22.5, 20, 17.5, and 15%, respectively. After 18 freeze-thaw cycles, the strength decreased by 2.66%, 11.85%, 18.83%, 16.79, and 29.02%, respectively. The predicted values of frozen soil binary medium model (BM), random forest model (RF) and PSO-LSTM model were compared with the measured values under different working conditions, and good accuracy was obtained. The R2 of the PSO-LSTM model test set was trained to more than 98%, and RMSE, MAE and MAPE were also trained to the lowest under the same working conditions. The influencing factors of deviator stress of frozen silty clay were given in order from strong to weak: initial moisture content>strain>confining pressure>number of freeze-thaw cycles. The LSTM optimal combination input parameters were searched by PSO, and the parameter adjustment speed of the model for the data learning process of frozen silty clay was greatly increased, which was conducive to the promotion of other soil constitutive prediction models. A new constitutive prediction model of frozen silty clay was developed using PSO-LSTM algorithm. 15 working conditions had been verified, and the optimal model had high accuracy in the constitutive prediction of frozen silty clay, which provided a good reference for the application of frozen soil engineering in cold regions.