Through questionnaire analysis, the cognitive situation and behavioral characteristics of different people around the nuclear power plant for nuclear emergency evacuation were obtained, and the possible behaviors of people during evacuation were predicted by Logistic model. Chi-square test and Logistic regression model were used to make correlation analysis and influence judgment on the evacuation psychology and evacuation behavior of different people in the area, and the optimal allocation algorithm was used to judge the prediction correctness of the model. The results showed that the factors such as whether to purchase nuclear safety-related insurance, the number of nuclear emergency evacuation drills, the understanding of nuclear emergency evacuation, and the presence of non-nuclear emergency protective measures at home had significant effects on the evacuation consciousness and psychology of the population. Gender, knowledge of nuclear emergency evacuation, the number of nuclear emergency evacuation drills participated, whether there are nuclear emergency protection measures at home and whether nuclear safety insurance has been purchased, and other factors have significant influence on crowd evacuation behavior. The simulation shows that people with certain nuclear evacuation experience can make relatively correct evacuation behavior and reduce evacuation time.