Abstract: Probabilistic integral method is a widely used method for predicting mining subsidence in China.How to accurately,quickly and reliably obtain the probability integral method model parameters based on the measured data has always been the difficulty of this method.Given this,based on the analysis of the advantages and disadvantages of WPA,this paper proposes the improvement strategy of second-migration mutation behavior,forms the improved wolves algorithm ( IWPA ),and introduces the IWPA into the probabilistic integral method for predicting parameters inversion,and constructs the probabilistic integral method model parameter inversion method (MIWPA) based on the improved wolves algorithm.The simulation results show that the relative error and median error of MIWPA inversion parameters are less than 3.4% and 4.02 respectively,and the accuracy and reliability of MIWPA are better than that of MWPA.Applying MIWPA to the parameter inversion of the probability integral method model of 1414(1) working face in Guqiao Mine of Huainan Mining Area,and the obtained probability integral model parameters were q=0.93,tan β=1.98,b=0.42,θ=84.53,S₁=-12.44 m,S₂=-18.80 m,S₃=55.06 m,S₄=33.98 m.The median error of subsidence and horizontal movement fitting was 114.88 mm,which met the engineering application accuracy.