With the development of technology, the technology and equipment for collecting physical signs are gradually maturing, and the data analysis ability and medical treatment technology are also developing rapidly. However, few models can analyze physical signs data in real-time and conclude, and then transmit them to emergency centers and hospitals. Therefore, a medical warning and automatic call for help model based on vital signs were constructed. Based on finite automata, the model took the analysis of physical signs data as the condition of state transition. To ensure the accuracy of the analysis of physical signs data, 448,972 data in the triage table of the MIMIC-IV database were used as data sets. Pycaret module was used to compare the models of various algorithms. Finally, the LightGBM algorithm was adopted, and supervised learning of data was carried out through parameter adjustment. The accuracy rate was more than 97%, indicating that the constructed model was feasible.