Crime Hot-Spot Modeling via Topic Modeling and Relative Density Estimation
This addresses crime analysis for law enforcement by providing a method to uncover unnoticed hot-spots, though it appears incremental as it combines existing techniques in a new way.
The authors tackled the problem of identifying crime hot-spots by analyzing crime record narratives using topic modeling and a novel kNN relative density estimation method, achieving results that captured geographic trends missed by dispatchers in a dataset of 475,019 documents.
We present a method to capture groupings of similar calls and determine their relative spatial distribution from a collection of crime record narratives. We first obtain a topic distribution for each narrative, and then propose a nearest neighbors relative density estimation (kNN-RDE) approach to obtain spatial relative densities per topic. Experiments over a large corpus ($n=475,019$) of narrative documents from the Atlanta Police Department demonstrate the viability of our method in capturing geographic hot-spot trends which call dispatchers do not initially pick up on and which go unnoticed due to conflation with elevated event density in general.