# Predicting Crime Using Spatial Features

**Authors:** Fateha Khanam Bappee, Amilcar Soares Junior, Stan Matwin  
**arXiv:** [1803.04474](https://arxiv.org/abs/1803.04474)  
**Date:** 2018-03-12  
**Summary page:** https://www.summarizepaper.com/en/paper/1803.04474v1/

## Key points

- Study aims to develop a machine learning model for crime prediction using geospatial features
- Utilizes reverse geocoding technique to retrieve open street map (OSM) spatial data
- Proposes identifying hotpoints extracted from crime hotspots area through HDBSCAN
- Computes a spatial distance feature based on the position of different hotpoints for various types of crime
- Tests engineered features using crime data obtained from Royal Canadian Mounted Police of Halifax, NS
- Results indicate that incorporating newly generated spatial features improves crime prediction accuracy
- Provides insights into utilizing geospatial data and machine learning techniques for effective crime prediction and prevention strategies

## Summary

The study conducted by Fateha Khanam Bappee, Amilcar Soares Junior, and Stan Matwin aims to develop a machine learning model for crime prediction using geospatial features. The researchers utilize the reverse geocoding technique to retrieve open street map (OSM) spatial data. Additionally, they propose identifying hotpoints extracted from crime hotspots area through the application of Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN). To further enhance their model, the researchers compute a spatial distance feature based on the position of different hotpoints for various types of crime. This computed value is then utilized as a feature for classifiers. The engineered features are tested using crime data obtained from the Royal Canadian Mounted Police of Halifax, NS. The results indicate that incorporating newly generated spatial features into the model significantly improves crime prediction accuracy. This research provides valuable insights into utilizing geospatial data and machine learning techniques for effective crime prediction and prevention strategies.