The purpose of this study is to support disaster prevention planning in underground shopping street by simulation of human behaviors during evacuation in a disaster. This study adopted Cellular automata(CA) as the technique analyzing evacuation behaviors. CA is interacting with neighbor cells in local. Evacuation behaviors are reproduced by comprehensively taking in various factors on evacuation behaviors using CA. Some improvements of the system were conducted considering beginning time of evacuation and the influence of a guide during evacuation. Developed program in this study can simulate evacuation behaviors caused by the fire occurred at a restaurant in Tenjinchikagai underground shopping street.