AbstractWhen a disaster such as a large earthquake occurs, the resulting breakdown in public transportation leaves urban areas with many people who are struggling to return home. With people from various surrounding areas gathered in the city, unusually heavy congestion may occur on the roads when the commuters start to return home all at once on foot. In this chapter, it is assumed that a large earthquake caused by the Nankai Trough occurs at 2 p.m. on a weekday in Osaka City, where there are many commuters. We then assume a scenario in which evacuation from a resulting tsunami is carried out in the flooded area and people return home on foot in the other areas. At this time, evacuation and returning-home routes with the shortest possible travel times are obtained by solving the evacuation planning problem. However, the road network big data for Osaka City make such optimization difficult. Therefore, we propose methods for simplifying the large network while keeping those properties necessary for solving the optimization problem and then recovering the network. The obtained routes are then verified by large-scale pedestrian simulation, and the effect of the optimization is verified.
Approx-SMOTE: Fast SMOTE for Big Data on Apache Spark