Forecasting seasonal criminality using SARIMA: an application to monthly aggravated assaults in California

TL;DR

This paper demonstrates that SARIMA models can effectively forecast monthly aggravated assaults in California up to six months ahead, aiding policymakers in resource allocation and crime prevention strategies.

Contribution

It applies SARIMA to model and forecast seasonal criminality, showcasing its effectiveness for state-level crime prediction using publicly available data.

Findings

SARIMA achieves reasonable accuracy up to six months ahead

Forecasts can inform resource planning and policy decisions

Seasonal patterns are significant in criminal activity data

Abstract

California experienced an increase in violent criminality during the last decade, largely driven by a surge in aggravated assaults. To address this challenge, accurate and timely forecasts of criminal activity may help state authorities plan ahead and distribute public resources efficiently to reduce crime. This paper forecasts monthly aggravated assaults in California using a publicly available dataset on state crimes and a time series SARIMA model that incorporates the highly seasonal behavior observed in the data. Results show that predictions with reasonable accuracy up to six months in advance are produced, showing the usefulness of these techniques to anticipate state-level criminal patterns and inform public policy.