Security Decision Governance Between Human Intuition and Mathematical Modeling: A Predictive Applied Study of Violent Misdemeanor Crimes in the State of Kuwait (1993-2024)

Abstract

This study examines the methodological challenge of transitioning from prediction to decision-making in police work through a critical and empirical comparison between intuition-based forecasting grounded in field experience and mathematical forecasting based on statistical modeling as a tool for evidence-based security governance. The study is based on the assumption that reliance solely on intuitive judgment is constrained by cognitive biases and limited reproducibility and verifiability. A quantitative analytical approach was employed using the Box–Jenkins methodology through an ARIMA model, based on the time-series data on violent misdemeanor crimes recorded in the State of Kuwait during the period 1993–2024. To identify the optimal forecasting model, several ARIMA specifications were compared using statistical criteria including Stationary R², RMSE, MAPE, and Normalized BIC. The findings demonstrated the superiority of the ARIMA (2,1,1) model as the best-fitting and forecasting model, achieving an explanatory power of 0.297 and a Mean Absolute Percentage Error (MAPE) of 10.4%. The study concludes that enhancing preventive security decision-making requires integrating professional expertise with statistical modeling within a data-driven governance framework to improve resource allocation efficiency and support proactive policing strategies