This research analyzes the group insurance cases which are still effective at the end of year 2007 of the ”S” Insurance Company. Logistic regression analysis is used to establish a discriminant model, and regression analysis is executed to find the important influential variables for compensation ratios. Research results show that, the accurate discriminant percentage of the logistic regression model whose independent variables are ”total insured number,” ”insure rate of medical insurance,” ”insure rate of fixed daily amount insurance,” ”insure rate of injure insurance” is 81.7%. After analyzing 16 independent variables and 2 area dummy variables, we construct several regression models for predicting the compensation ratios, and suggest insurance companies need to have different regulations for insuring different group insurance projects.