Abstract: With the economic development, the number of commercial complexes in major cities is increasing, and shopping and entertainment in commercial complexes are favored by consumers. Due to the large building area and complex internal structure of the commercial complex, in case of fire, most people are not familiar with the safety exit path, which is not conducive to evacuation. In order to study the evacuation psychology and behavior of people in the event of a fire, this study used Blender and Unity 3D to construct a VR fire scene of a large commercial complex, combined with the monitoring of subjects' heart rate change, escape time and the correct escape, carried out K-means clustering analysis on the experimental data by machine learning. The purpose of this study is to classify and analyze the evacuation psychology and behavior of personnel, and at the same time to enhance the evacuation safety awareness of personnel and improve the safety education of fire evacuation ability.

Key words: VR technology, machine learning;K- means cluster analysis