Using tweets to understand changes in the spatial crime distribution for hockey events in Vancouver. (25th April 2018)

• Record Type: Journal Article
• Authors: Ristea, Alina; Andresen, Martin A.; Leitner, Michael
• Is Part Of: Canadian geographer. Volume 62:Number 3(2018)
• Page Start: 338
• Page End: 351
• Publication Date: 2018-04-25
• Subjects: spatial crime analysis -- Twitter -- hockey -- geographically weighted regression; analyse spatiale de la criminalité -- Twitter -- hockey -- régression pondérée géographiquement; Geography -- Periodicals; 910
• DOI: 10.1111/cag.12463
• Languages: English
• ISSNs: 0008-3658

Abstract

Abstract: The use of social media data for the spatial analysis of crime patterns during social events has proven to be instructive. This study analyzes the geography of crime considering hockey game days, criminal behaviour, and Twitter activity. Specifically, we consider the relationship between geolocated crime‐related Twitter activity and crime. We analyze six property crime types that are aggregated to the dissemination area base unit in Vancouver, for two hockey seasons through a game and non‐game temporal resolution. Using the same method, geolocated Twitter messages and environmental variables are aggregated to dissemination areas. We employ spatial clustering, dictionary‐based mining for tweets, spatial autocorrelation, and global and local regression models (spatial lag and geographically weighted regression). Findings show an important influence of Twitter data for theft‐from‐vehicle and mischief, mostly on hockey game days. Relationships from the geographically weighted regression models indicate that tweets are a valuable independent variable that can be used in explaining and understanding crime patterns. Key Messages: Spatial concentrations of crime have different patterns for hockey game and comparison days in Vancouver, Canada. Local and global regression models show crime‐related tweets are a better explanatory variable for crime than all tweets. Specifically, two of the six crime types develop a stronger connection with crime‐related tweets, namely theft‐from‐vehicle and mischief.