Objective To analyze the potential categories of the recognition and response pattern of stroke warning signs in young and middle-aged people, and explore the characteristics of different categories, so as to provide the scientific basis for personalized education in the future. Methods A self-designed questionnaire in terms of recognition of and response to the stroke warning signs were administered to 564 young and middle-aged people. Based on their recognition of and response to the stroke warning signs, latent class analysis was run to generate the latent classes of the people. Then multinomial logistic regression analysis was perform to probe effects of demographic-sociological information, the stroke experience of relatives and friends and the knowledge of "stroke 120" on the latent classes. Results A 3-class model was the most preferable one, with the latent classes being named "high recognition of stroke warning signs-call 120 group" (19.1%), "sensitive to physical symptoms-hospital group"(52.3%), "overlook limb numbness-dislike hospital group" (28.6%) respectively. Those who worked in service sector, aged 51-60, who had relatives and friends with stroke experience, or who had knowledge of "stroke 120" more likely belonged to the "high recognition of stroke warning signs-call 120 group" (P<0.01,P<0.05). Conclusion The results highlight the key points of health education for various groups of people. To improve teaching effect, it is suggested to provide targeted education tailored to the characteristics of different groups of people in the future.