Abstract: This paper, by adopting the methods of questionnaire, statistics and mathematical modeling, designs and builds a physical fitness risk pre-warning model for the civil servants in Shanghai.Based on the cluster analysis and stratifying analysis of the sample data, it then explores the deviation both individually and collectively.The results show that the single-indicator risk assessment and the general-indicator risk assessment of the civil servants' physical fitness have a high goodness of fit under the condition of gender clustering and age stratifying, with an obvious group deviation though. Besides, there is an ideal distribution of the large sample's risk pre-warning value with a good convergence, which can effectively describe civil servants' health risk pre-warning degree, showing the effectiveness of this pre-warning model.