Abstract: A geological model classification and step-by-step modeling method is proposed to address the issue of differences between existing mining prediction software calculation models and actual geological conditions, as well as the inability to predict uneven subsidence caused by difference in internal thickness and undulation of mining blocks. The traditional probability integration algorithm and formula are optimized based on the characteristics of differences in coal thickness and dip angle of each mining differential unit underground, with the aim of building a framework first and then interpolating the details. On this basis, a software for predicting mining subsidence was developed, and the Zhaolou Mine 5301 working face was taken as an engineering example to compare it with existing algorithms and optimization algorithms. After analysis, it was shown that the optimized probability integral method prediction algorithm can reflect the uneven subsidence caused by differences in coal thickness and other factors, and the prediction accuracy has been improved.