Abstract:
A natural disaster is an event which causes damage to both lives and properties. The
detection of natural disasters is an important issue. Social media is a powerful
source which can be used to improve managing disaster situations. Post-disaster
management can be largely improved by proper social media mining since social
media are capable of sharing information in a real-time manner. After identifying
the importance of social media for post disaster management, Twitter posts were
fetched from the Twitter API using predefinedKeywords relating to the disaster. At
the second stage these posts were cleaned and the noise was reduced. Two-level
filtering of non-relatedKeywords was used. Then at the third stage, the geolocation
and the disaster type was identified. The Named Entity Recognizer library and the
Google Maps Geocoding API were used to obtain the geolocation. The same three
steps were carried out for the news, which was fetched from the News API. Finally,
each datum from the Twitter API was compared with the relevant datum from the
News API to give a rating for the trueness of each post. The rating of “More accurate”
was obtained by 24% of the posts. The ratings of “Moderately accurate” and “Less
accurate” were obtained by 15% and 13% of the posts respectively. Remaining 48%
posts obtained the rating of “No correlation”. This model can be used to alert
organizations to carry out their activities of disaster management in a timely
manner. The future development steps are as follows: to integrate the other social
media to fetch data, to integrate the weather data into the system in order to
improve the precision and accuracy for finding the trueness of the disaster and
location and the use of some sophisticated machine learning techniques to reduce
the noise.