Georeference Social Sensing for Disaster Response Assessment using Support Vector Machine

Abstract

Social sensing is based on the idea that communities or group of people can provide a set of information analogous to those what we can achieve from a sensor network. Classifying this considerable information, produced during and after the disaster could significantly help the government in making an informed situational assessment for relief operation. Support Vector Machine (SVM) was used to classify tweets from typhoon Melor using a tf-idf as an implementation of a bag of words model for data representation. The cleansed data were used to train the SVM following a five-fold cross-validation technique, geolocation referencing from the tweets were used to obtain location. The resulting corpus was plotted on the map as an assessment tool, that would be a valuable tool for disaster management.