Abstract: In order to further improve the accuracy and efficiency of underground fault identification, an intelligent fault recognition model based on the extreme gradient boosting tree (XGBoost) machine learning algorithm was constructed for coal seam faults, combined with the particle swarm optimization (PSO) algorithm to optimize the model's related parameters. A forward model was established to verify the PSO-XGBoost model, and the classification prediction performance of the PSO-XGBoost model was compared with that of the PSO-RF and PSO-SVM models based on actual data collected from the Diandong mining area. The accuracy rate and log loss value were selected as the main evaluation indicators to evaluate the accuracy of the classification prediction models for each model. The results show that the PSO-XGBoost model has a high accuracy in fault structure identification; the PSO-XGBoost model has higher accuracy and better stability in fault identification.