For the data-processing reliability problems in risk-assessment group decision-making, the importance of the expert weights for the data processing was proposed. From the industry characteristics of automotive crash safety research, based on expert social factors indicators, by using two analysis methods of system clustering and K-means clustering, clustered the experts according to expert-authority-weight; Based on the data consistency and the consistency degree of individual and group from expert opinions, by using the method based on information entropy, clustered the experts according to expert-opinion-weight; Two clustering results were compositely clustered to get the final weight of the experts. The results show that the expert weights got by the composite clustering have good proportionality and reliability, based on compositing expert-authority-weight and expert-opinion-weight, it lays a foundation for the study of systems projects.