Detecting Flood Vulnerable Areas in Social Media Stream Using Association Rule Mining

Abstract

In this study, we identify flood vulnerable areas by employing association rule mining to social media streams. The following processes are involved: (1) data collecting; (2) data cleaning; (3) representing the training data; (4) determining the association between words; and (5) using the association values as guide to identify vulnerable areas. As testbed, we focused on tweets from Metro Manila, particularly tweets from August 2015. We decided to use tweets since it is publicly available. This study will aid different government agencies, specifically those that are focusing in disaster management and others that are into flood related proj ects. This paper presents the possibility of detecting location in Metro Manila, which in turn gives higher possibility of being able to trace possible flood vulnerable areas. As future works, since the entity extraction is done manually an automation of this can be very helpful to other researchers.