A Support Vector Clustering Based Approach for Spatiotemporal Analysis in Security Informatics

Security informatics involves the application of information technology to protect public health and security. A common research topic in security informatics is the identification and representation of clusters of events (e.g., a disease cluster or a crime hot-spot). Understanding why clusters change shape and move over time would be valuable to researchers in security informatics, providing them with greater means for discovering causes as well as examining the effectiveness of mitigation efforts. A first step towards this type of understanding necessitates the establishment of methods for the description of how clustering events evolve over time. However, existing approaches for the analysis of clusters are limited in their ability to describe spatiotemporal behaviour such as movement and deformation. This research presents a framework for facilitating such spatiotemporal descriptions based on support vector clustering and the spatiotemporal helix. Benefits of this approach include the absence of bias a prioi regarding the shape or number of clusters and the ability to describe spatiotemporal behaviour in terms of both changes in shape and movement. Results based on simulated data suggest the effectiveness of this approach for spatiotemporal analysis in a range of application domains in security informatics. * Corresponding author.