The pressure condition of fully mechanized mining face is an important reference factor for formulating the support design of mining engineering. The accurate prediction of the pressure trend of fully mechanized mining face can not only assist designers to make reasonable underground support scheme, but also be of great help to the safe and efficient production of coal mine. In this paper, a model based on PSO-GBDT is integrated to predict the first pressure of near shallow coal seam. Eight integrity-good factors which have significant influence on the pressure of fully mechanized mining face in near shallow coal seam of Datong mining area are selected as features. The first pressure step distance and strength are taken as output, combined with particle swarm optimization algorithm. GBDT is used to construct the prediction model of the first pressure in near shallow coal seam. Visual data analysis was carried out on the data set first, and then to GBDT integrated learning method to build the first pressure prediction model, and then by particle swarm optimization algorithm for trained GBDT super parameter optimization, and finally to the various characteristics to build the prediction model of influence weight sorting and provide the reference for pressure mechanism analysis, And compared with GBDT or Adaboost manually modeling method. The experimental results show that PSO-GBDT has achieved better results in predicting the first pressure step distance and strength of near shallow coal seam, the performance evaluation indexes of begin{document}$ mathrm{R}mathrm{M}mathrm{S}mathrm{E} $end{document} and begin{document}$ {{R}}^{2} $end{document} have achieved better results, compared with the GBDT manually modeling method, the begin{document}$ {{R}}^{2} $end{document} indexes of PSO-GBDT are improved by 7.2% and 6.8% respectively, and the begin{document}$ mathrm{R}mathrm{M}mathrm{S}mathrm{E} $end{document} indexes are reduced by 39% and 42% respectively, in the model constructed with the first pressure step distance and strength as the output. In addition, the comparative experimental results of PSO-GBDT and Adaboost manually modeling methods show that the PSO-GBDT has significant advantages in prediction performance. The method calculates the first pressure distance and strength of fully mechanized mining face with the characteristics of high performance and multi-factors, which provides a new solution for accurately predicting the first pressure trend of fully mechanized mining face.