The Utility of Point of Interest Data for Crime Risk Prediction
Paweł Cichosz , Zbigniew M. Wawrzyniak , Radosław Pytlak , Grzegorz Borowik , Eliza Szczechla , Paweł Michalak , Dobiesław Ircha , Wojciech Olszewski , Emilian Perkowski
AbstractThis paper examines the utility of Point of Interest (POI) data for learning crime prediction models. Crime event locations for a Polish city are aggregated into a grid and merged with selected data layers from OpenStreetMap. The resulting dataset contains crime count attributes and POI count attributes for particular areas of the city, obtained by aggregating over a rectangular grid. After identifying high-risk areas based on crime counts, POI count attributes are used for learning crime risk prediction models with the logistic regression, support vector machines, and random forest algorithms. The experimental results suggest that POI attributes have high predictive utility. Classiffcation models using these attributes, without any form of location identiffication, exhibit very good predictive performance, which makes it possible to reuse them over different cities.
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.