Spatio-temporal stochastic differential equations for crime incidence modeling.

Area of Science:

• Quantitative analysis
• Spatio-temporal modeling
• Stochastic processes

Background:

• Real-world crime data presents complex spatio-temporal patterns.
• Accurate forecasting of criminal incidents is crucial for public safety and resource allocation.
• Existing models may not fully capture the intricate dynamics of crime distribution.

Purpose of the Study:

• To develop and validate a methodology for quantitative fitting and forecasting of real spatio-temporal crime data.
• To model the annual-trend components of monthly crime time series within specific geographic areas.
• To identify high-risk areas and periods for crime by analyzing historical data.

Main Methods:

• Utilizing stochastic differential equations for data analysis.
• Applying Itô diffusion to model annual-trend components of crime time series.
• Incorporating jointly correlated noises to represent district-level crime interdependencies.

Main Results:

• Successfully fitted and forecasted spatio-temporal crime data for Valencia, Spain (2010-2020).
• Modeled monthly crime incidents across 26 zip codes using Itô diffusion.
• Demonstrated the capability to account for correlated noise patterns between districts.

Conclusions:

• The proposed methodology provides a robust framework for analyzing and predicting crime patterns.
• This approach can enhance situational awareness by identifying potential crime hotspots.
• The findings support data-driven strategies for crime prevention and management.