Behavior Support for Residents During Typhoon Based on Association Rule Mining

This study aims to develop an information system for providing decision support with regard to the behavior of residents during a typhoon. Typhoon disasters have relatively longer duration than earthquake disasters; therefore, the reinforcement of preventative measures can reduce the disaster's magnitude. In the event of a disaster, alarms are initiated and evacuation information is provided by disaster management organizations; however, human casualties and material damage still occur because of delays in taking appropriate action. For this reason, action trigger information is required. Such information will facilitate evacuation through processes familiar to the public. In this study, action trigger information for residents was extracted from damage data of past typhoon disasters. Action triggering information consists of typhoon-related damage occurrence relationships, which are established using association rule mining. Furthermore, an information system was developed for the purpose of supporting the behavior of residents in real-time, during the typhoon, by using the generated association rules and microblogging content. Several experiments were performed to confirm the validity of association rules and their applicability to real-time forecasting.