In Japan, tsunamis and landslides are becoming more common as secondary disasters after earthquakes, and early evacuation is an important issue to reduce the damage. In the previous study, the authors proposed a method to shorten the time to complete evacuation using social graphs and to reduce the number of messages in evacuation based on geographical information. However, it did not consider the weight of links among social graphs and ignored the relationship among evacuation targets (nodes). Therefore, in this paper, the authors improve the conventional method by adding weights among social graphs. In addition, the authors created a social graph with a bias in the placement of people, and evaluated whether the evacuation completion rate and time change depending on the bias in the placement of nodes.