Modelling underreported spatio-temporal crime events

Area of Science:

• Criminology
• Data Science
• Public Policy

Background:

• Crime statistics are crucial for public security strategies but are affected by underreporting, creating a 'dark figure of crime'.
• Accurate crime incidence and underreporting rates are essential for effective policy design.

Purpose of the Study:

• To develop and validate a novel methodology for estimating true crime incidence and underreporting rates over time.
• To address the challenge of the 'dark figure of crime' using sequentially available daily data.

Main Methods:

• A novel spatiotemporal event underreporting model based on the combinatorial multi-armed bandit framework was proposed.
• The model's parameters, including true incidence and underreporting rates, were validated through extensive simulations.
• The methodology was applied to real-world crime data from Bogotá, Colombia.

Main Results:

• The proposed methodology successfully identified fundamental model parameters in simulations.
• Estimated true crime and underreporting rates were derived from Bogotá's crime data.
• The study demonstrated the potential for rapid estimation of underreporting rates for spatiotemporal events.

Conclusions:

• The developed methodology offers a promising approach to accurately estimate crime underreporting.
• This technique can significantly aid public policy design by providing more reliable crime data.
• Addressing the 'dark figure of crime' is critical for enhancing citizen security.