Crowdsourcing Disaster Response: Application to enable proactive risk mitigation in the UAE

Abstract

As urban populations continue to grow, the potential for damage caused by natural or man-made disasters increases. During a crisis, timing is crucial. Anticipating possible outcomes and quickly alerting people can drastically reduce economic and social impacts. At the same time, social media is currently playing a prominent role during crisis around the world. Social media companies such as Facebook are developing features to help victims communicate with their loved ones during disasters. Also, crowdsourcing platforms are arising as a way to decrease costs of gathering and validating data. The aim of this thesis is developing an online crowdsourcing platform using social media for monitoring disasters in the UAE. Moreover, It will use machine learning techniques to filter, validate and broadcast information in real-time during a crisis. Also, we have developed a framework that enhances the solutions proposed in the literature. We have found that low reliability and unstructured data from social media can affect the performance of this system. The platform developed addresses this issue using crowdsourcing validation. It incentivizes users by gamifying the user experience on the platform. Finally, we have gathered a dataset of social media posts and developed a machine learning model to filter data coming from this source.

Date of Award	May 2017
Original language	American English