National Security through Social Media Intelligence: Domestic Incident Prediction

Abstract

Social media is the most responsive communication method on the Internet. It has attracted people from various backgrounds, including older and younger people. Social media is also a robust network system that connects users irrespective of their geographical locations. It has enabled people to express their views to more audiences than ever before. The database collections of online conversations on different topics may include public debated issues, supports, politics, religions, and tribal ideologies. There is no constriction to the type of information that users can share on social media platforms. However, due to the exceptional influence and reach of social media platforms, they have become tools in the hands of individuals and groups for different crimes ranging from stalking, cyberbullying, group coordination of riots, terrorist groups, for recruiting, radicalizing new members, raising resources, organizing activities, and exchanging information on how to trigger attacks in public places. Hence, an automated study of social media dataset comments will help to predict and solve security problems. This study considered the machine learning techniques on Twitter datasets to classify tweets as mass-shooting-incident tweets or non-mass-shooting-incident tweets. The decision tree, logistic regression, and support vector machine algorithms were combined to determine the accuracy of the machine learning approach in predicting mass shooting incidents.