Disaster Detection Using Twitter Data

How To Run

python ./detect.py <file containing tweets>

You can also type python ./detect.py -h for more help on running the program

How To Test The Classifier

The classifier can be tested using the command: python classify.py --test

This will test the classifier using a 90% training to 10% testing split and average the results over 10 interations.

Data

• Hurricane Irene tweets are available at http://apollo.cse.nd.edu/datasets/irene_hurricane.txt
• Hurricane Sandy tweets are available at http://apollo.cse.nd.edu/datasets/sandy_hurricane.txt
• The tweets collected from Twitter streaming api are available in the data folder

Plotting the Relevant Tweets Per Hour

A csv file containing the number of relevant tweets per hour chronologically can be generated using the following command:

python plot.py <data>

This will output the resulting csv file to the distributions directory.

Requirements

• python 2.7
• scipy
• numpy
• sklearn