CalGIS2016

Social media is the de facto communication apparatus to disseminate one’s self revelations. These profound disclosures often contain double-talk, peculiar insights, or contextual information about real-world concepts. Natural language processing is regularly used to uncover both obvious and latent knowledge claims within disclosures submitted amid such a complex environment. For example, a bystander may have first-hand knowledge of criminal incident, and using social media publish critical information about it. A geographic information system (GIS) is capable of large-scale point data analysis and possesses methods that enable processing, evaluation, and automatic spatial visualization of its dataset. Such artifacts when fused with traditional environmental criminology theory and social media erect guidelines, tools, and models for substantive construction and evaluation of GIS crime analysis solutions. On condition that a social media stream is timely and correctly processed, corrective action can be taken.

The construction of a natural language processing social media annotation pipe, to identify latent indicators extracted from a social media corpus, was an integral part of societal mishap prediction. Spatial visualizations and regression analyses were used to describe and evaluate project artifacts. As a result, social media corpora was operationalized; and subsequently used as a proxy for a traditional environmental criminology risk layer in construction of a social media GIS crime analysis artifact. Using such multi-domain collaboration, the artifact was able to increase the predictive crime incident outcome with an overall R-squared increase of 7.26%. This result is the state-of-the-art; there are no other results to compare it to.

Presenter: Anthony Corso, Ph.D., M.B.A., Assistant Professor, California Baptist University
Riverside, CA