Criminal cross correlation mining and visualization

Abstract

Criminals are creatures of habit and their crime activities are geospatially, temporally and thematically correlated. Discovering these correlations is a core component of intelligence-led policing and allows for a deeper insight into the complex nature of criminal behavior. A spatial bivariate correlation measure should be used to discover these patterns from heterogeneous data types. We introduce a bivariate spatial correlation approach for crime analysis that can be extended to extract multivariate cross correlations. It is able to extract the top-k and bottom-k associative features from areal aggregated datasets and visualize the resulting patterns. We demonstrate our approach with real crime datasets and provide a comparison with other techniques. Experimental results reveal the applicability and usefulness of the proposed approach.