A Framework For Profiling Crime Reported Using Social Media – A Case Of Twitter Data In Kenya

Crime profiling helps law enforcement agencies understand, tackle and sometimes predict the next move by criminals. This can be achieved by monitoring and studying patterns and trends that have occurred in the past and continue to occur in the present. Social media platforms such as Facebook, Google Plus, Instagram, Reddit and in this case Twitter, have created platforms where people share views, opinions and emotions all the while influencing and informing others.

This research set out with four objectives that would enable it to be successful in coming up with a framework for profiling. The objectives were, Identifying and extracting crime data from social media, Performing Sentiment analysis on crime data, Designing a Frame work for profiling of crime, using the Twitter social media platform and Testing of the framework.

This study set out to find crime related data from members of the public from Twitter and see if it can be used to profile crime. Collecting and extracting of tweets was done from the Kenyan Twitter population, cleaned and sentiment analysis carried out. The analysis of sentiments was done using a lexicon based algorithm within the SentiStrength open source tool. 

Through the framework model created in the project, the preprocessed data was analyzed in regards to the environment and author variables. Then establishing how they can be used to monitor or profile crime. The findings from this analysis led to viewing of patterns and hotspots of crimes on real life digital maps.

The results provided insight into the dynamics around different authors who tweet about crime and what differentiates one from another. These results where then investigated and compared to derive an understanding of the dynamics. Finally the framework was tested via two types of tests, usability test and Pearson correlation coefficient.

The major study findings showed that it is possible to come up with a framework model to use in the tracking of crime in Kenya from Twitter. Usability and correlation coefficient tests proved the framework successful from user feedback and remodeling of the framework.

Major recommendations for future studies is to have a data collection method for a streaming API instead of a standard API. This API together with a tailor made application that collects real time data will ensure better collection of data. The Law Enforcement Agencies can then collect data continuously and store in their databases. There is also a need to have local language content within the dictionaries. This will better localize the sentiments of the tweets to include few Swahili, Sheng or tribal words.