In order to improve the prediction precision and application values of evacuation time model (briefly called an empirical model) that is generally accepted by the whole world, the influences of individual attributes (including both natural and social attributes) are analyzed. Based on regression analysis and the evacuation investigation on a large university gymnasium, a 185-set-data sample is established. Combined with the factors that affect evacuation behavior and time of individuals, the regression model of crowd evacuation time is built with the statistical software, SPSS 17.0. The individual attributes are consolidated into crowd attributes and the correction coefficient 'k' is solved on the premise of regression model. The empirical model is revised and examined through the utilization of correction coefficient 'k'. Finally, a fire evacuation design for a university stadium project is analyzed based on the optimized empirical model. The results indicate that with the use of regression analysis and correction coefficient 'k', the weakness of empirical model involving lack of consideration on the influences of crowd attributes could be effectively overcome, and both the prediction accuracy and application values of evacuation time model would be improved.