Interaction of Crime Risk across Crime Types in Hotspot Areas

Hong Zhang, Yongping Gao, Dizhao Yao, Jie ZhangISPRS Int. J. Geo Inf 2023 · DOI

TL;DR

This paper investigates how crime risk spreads between different crime types within urban hotspots, moving beyond the established understanding of repeat victimization within a single crime type. By analyzing crime data from Suzhou, China, the researchers developed a novel clustering algorithm to identify stable crime hotspots and then used Pearson correlation to map the interactions of crime risk across various crime categories, offering insights for more targeted crime prevention strategies.

What Problem Does This Paper Address?

While the concept of "repeat victimization" – where a victim is targeted again for the same type of crime – is well-established, and the idea of "near-repeat victimization" (crimes occurring in close proximity in time and space) is also understood, there's a significant gap in understanding how crime risk might transfer across different crime types. Criminals might leverage opportunities or information gained from one type of crime to commit another. This paper aims to fill this gap by exploring the "interaction of crime risk across crime types in hotspot areas." Understanding these cross-crime type dynamics is crucial for developing more effective and nuanced crime prevention strategies that go beyond addressing single crime categories in isolation.

Methods

The study employs a two-pronged approach:

  1. Hotspot Extraction: The researchers propose a novel algorithm called KMNN-DBSCAN (K-means nearest neighbor DBSCAN). This algorithm is an enhancement of the traditional DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm. The key innovation is its ability to automatically adapt the DBSCAN parameters (Eps and Minpts) based on the dataset's distribution, overcoming the limitations of manual parameter tuning which can be sensitive to density variations. This allows for the detection of stable crime hotspots with varying densities.
  2. Crime Risk Interaction Analysis: Once stable hotspots were identified, the study utilized the Pearson correlation coefficient to quantify the relationship between different crime types. Specifically, they analyzed the correlation between the number of cases of one crime type in a given week and the number of cases of another crime type in the following week. This temporal lag is crucial for understanding the spread of risk.

Datasets

The study uses crime data from Suzhou City, China, specifically from four districts (Huqiu, Gusu, Wuzhong, and Xiangcheng) covering the period from December 2011 to October 2013. The dataset includes weekly counts for ten typical crime types: pickpocketing, vehicle property theft, office creeper activity, vehicle theft, shoplifting, telecom fraud, robbery, home burglary, internet fraud, and delict.

Key Insights and Ideas

  • Crime risk isn't isolated: The study confirms that crime risk can indeed spread across different crime types, not just within the same category. This suggests a more interconnected criminal ecosystem within hotspots.
  • Hotspot stability matters: The research emphasizes the stability of crime hotspots, meaning these areas consistently attract criminal activity over time. This stability provides a reliable basis for analyzing crime interactions.
  • "Flag" and "Boost" concepts extend: The findings align with existing theories like "flagging" (marking targets) and "boosting" (providing assistance), suggesting these mechanisms can operate across different crime types. For example, information gained from observing potential targets for one crime might be used for another.
  • Opportunity drives cross-crime behavior: The "Routine Activities Theory" and "Rational Choice Theory" are supported, as criminals appear to exploit available opportunities and make choices based on perceived risks and benefits, which can involve shifting between crime types.
  • Specific crime pairs show strong correlations: Certain crime types exhibit notable positive or negative correlations, indicating specific pathways for risk transfer (e.g., vehicle theft and home burglary, or shoplifting and office creeper activity).

Key Results

  • The KMNN-DBSCAN algorithm successfully identified nine stable crime hotspot regions in Suzhou.
  • The analysis revealed significant correlations between various crime types within these hotspots. For instance:
  • Hotspot 2: Strong positive correlation between vehicle theft and robbery (0.775), and robbery and home burglary (0.605). A strong negative correlation was found between robbery and vehicle theft (-0.756).
  • Hotspot 5: Strong correlations between office creeper activity and vehicle theft (0.618), and shoplifting and telecom fraud (0.632).
  • Hotspot 9: Strong correlations between vehicle property theft and pickpocketing (0.697), vehicle property theft and burglary (0.625), and robbery and shoplifting (0.651).
  • Conversely, some crime types showed negative correlations, suggesting they might be mutually exclusive or that the conditions favoring one deter the other (e.g., shoplifting and office creeper activity in Hotspot 5).
  • The study found that in areas with high crime rates and diverse land use (like Gusu District), correlations between crime types were not always strong, suggesting a degree of independence or that each crime type might be relatively independent in such environments.

Limitations and Future Work

  • Data limitations: The study acknowledges that the detection rate for certain crimes, like telecommunications fraud, is extremely low due to the complexity of criminal networks. The analysis is based on reported crimes, which may not capture the full extent of criminal activity.
  • Temporal resolution: While a weekly interval was used, finer temporal resolutions might reveal more nuanced interactions.
  • Geographical scope: The study is limited to Suzhou City. Generalizing these findings to other urban environments would require further research.
  • Future work: The authors suggest that understanding these correlations can help police officers develop more effective crime prevention programs by predicting potential offender behavior and helping victims reduce their vulnerability. They also highlight the need for continued efforts in anti-fraud propaganda and strengthening law enforcement systems, particularly for emerging crime types like telecommunications fraud.

Summary updated

BibTeX
@article{zhang2023interaction,
  author = {Hong Zhang and Yongping Gao and Dizhao Yao and Jie Zhang},
  title = {{Interaction of Crime Risk across Crime Types in Hotspot Areas}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2023},
  doi = {10.3390/ijgi12040176},
}
Figure