aolmez / terroristtracker

TerroristTracker

This software is a portion of the algorithm used for the research project by Mihir Patel and Nikhil Sardana. The uploaded code includes parts of the image recognition program, text analysis and overall evaluation networks, but is not complete as certain parts of the code will remain private. Some parts may be outdated as this is a code dump.

The purpose of the algorithm is to automatically process and detect terrorist accounts on social media platforms. This is accomplished by scanning profiles for ISIS flags using deep learning (convolutional neural networks), analyzing key words, and making predictions based off of user statistics. Through testing, it was found to have an accuracy of 93.5% in detecting flags and an accuracy of 90.5% in classifying users. More recent models have reached 97% accuracy in image detection and 93.5% overall.

This project won Grand Prize at Intel Regional Science Fair and was selected as Intel International Science Fair Finalist. The research has been presented to organizations like the NSA and Raytheon.