Crime Hot-Spot Modeling via Topic Modeling and Relative Density Estimation

TL;DR

This paper introduces a novel approach combining topic modeling and relative density estimation to identify crime hot-spots from narrative data, revealing hidden spatial patterns overlooked by traditional methods.

Contribution

The paper proposes a new method that integrates topic modeling with kNN-based density estimation to detect crime hot-spots from narrative texts, improving spatial trend detection.

Findings

Effectively captures geographic hot-spot trends.

Identifies hidden patterns not detected by initial dispatch analysis.

Demonstrates scalability on large crime record datasets.

Abstract

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.