Abstract: In earthquake prediction work, mining group characteristics under the large data and characterizing it by some proper parameters was necessary. The authors defined the group characteristic parameters,based on which the mathematical model of seismic damage cumulative probability was proposed; Secondly, an average parameter representing sample unit group characteristics, whose reliability could be tested by being compared with the sample damage result,was obtained within a certain range precision by regression analysis and the iterative calculation of sample unit data; Finally,these parameters were used in the calculation of target cell damage model,and the corresponding predictive result was obtained. Meanwhile,the authors verified the reliability and feasibility of the method by an instance.