Toward Predictive Crime Analysis via Social Media, Big Data, and GIS Spatial Correlation

Abstract

To support a dissertation proposal a link between social media, incident-based crime data, and data of public domain needed to be verified. A predictive crime-based artifact utilizing data mining and natural language processing techniques commingled with graphical information system architecture is complex. With respect to social media, an attempt was made to observe such an artifact’s data flexibility, process control, and predictive capabilities. Data and their capabilities were observed when preprocessing social media’s noisy data, government-based structured data, and obscurely collected field data for use in a predictive GIS artifact. To support project goals the approach for artifact design, data collection, and discussion of results was couched as an exploratory study. Results indicate a link between social media data and domain specific datasets exist. Questions for further observation and research deal with processing the subtle differences between structured and noisy data, weighted social media input layers, and time-series analysis.