Spatial crime distribution and prediction for sporting events using social media

Area of Science:

• Criminology
• Data Science
• Urban Studies

Background:

• Sporting events draw large crowds, increasing social media usage.
• Research suggests a link between sporting events and increased crime rates.
• Geospatial analysis and social media data offer new avenues for crime research.

Purpose of the Study:

• To analyze the spatial correlation between crime and geo-located Twitter data during sporting events.
• To develop and evaluate crime prediction models incorporating social media information.
• To compare crime patterns on game days versus non-game days for basketball and hockey.

Main Methods:

• Spatial analysis of crime occurrences, demographic, socio-economic, and environmental variables.
• Analysis of geo-located Twitter messages, including a subset of 'violent' tweets.
• Application of feature selection and localized kernel density estimation for crime prediction.

Main Results:

• A significant spatial correlation was found between crime and Twitter activity, especially 'violent' tweets.
• Twitter data, including violent tweets, demonstrated utility in enhancing crime prediction models.
• Crime prediction models showed varying levels of improvement for different crime types on game and non-game days.

Conclusions:

• Geo-located social media data, particularly concerning violent content, can be a valuable covariate in crime prediction models.
• The study highlights the potential of integrating real-time social media data into law enforcement and urban planning strategies.
• Findings are applicable to predicting crime related to major sporting events and general urban safety.