Abstract
In order to secure safe railroad network from rainfall-induced natural disaster such as landslides, each railway company in Japan has own regulations concerned with train operations and district patrols for such conditions. Recently more detailed information on rainfall has been required to meet the increasing frequency of train services. For instance, adopting ten-minute rainfall data instead of conventional hourly rainfall data is in urgent need for the revision of the regulations, especially when the case of concentrated heavy rainfalls is targeted; however the data to be treated would become larger and more complicated. Thus, this paper proposes an application of “support vector machine”, which enables quite rapid and precise analyses of huge amount of ten-minute rainfall data, as a sequential renewable method of the train operation regulations.
