The high profits posted each year by one of the major mobile phone operators in Kenya is an indication of the high smart phones and data usage in the country. Further the highest percentage of this mobile data usage is on social media which includes facebook, twitter, whatsapp, instagram, snapchat and others. While a lot of organizations especially companies have leveraged on information shared on these platforms to improve on product and service delivery while fostering better customer service and stakeholder relations, very little has been done on access and use of crime information shared on social media platforms such as twitter.
In light of this, the project set out to leverage on social media, specifically twitter, in Nairobi, Mombasa and Kisumu through three main objective. The first objective was to establish how available social media data can be used to map crime. Illustrations through literature were made on how other implementations have taken advantage of this information to achieve set objectives. The second objective was the development of a tool which enabled the collection and processing of the twitter information to facilitate mapping of crimes to provide awareness.
The third objective was to test the developed crime mapping tool to ensure that it performed as intentioned using the correct data. Descriptive research was used during the study. This is because available existing data on twitter was used. This data was collected from twitter through a twitter API integrated in a java program which abets in specifying the information of key interest to be collected. Information shared on twitter is known as tweets. With millions of tweets been tweeted each day on different topics, it is crucial to be able to only pick what is of interest. Once this information on the area of interest was mined, further filtering by use of a natural language processing tool was done to determine the true meaning of the tweet through a process known as sentiment analysis. Tweets that were found to imply that crime was committed were sent and stored into a database. Data analytics were performed on the data to identify patterns for mapping purposes. The Data analytics was done using WEKA, an open source data mining tool. The complete model was then tested to measure its accuracy.
The developed crime mapping tool was found effective in collection of the twitter data as specified. However the Stanford CoreNLP sentiment analysis used tool was not a 100% accurate. This is having classified some tweets as denoting crime while through a manual analysis no crime actually happened. Another key finding through the interviews conducted is that despite the many twitter users in Kenya, very few used the twitter platform to share information about crime.
The major recommendations were the use of a hybrid sentiment analysis tool in future works to improve accuracy and encouragement of the use of twitter to share information about crime. Also the use of a web-based crime mapping tool integrated with a map application to widen accessibility and better visualization options.
