Using Some Artificial Intelligence Algorithms to Estimate the Poisson Regression Model

Abstract

This research aims to estimate the Poisson regression model to study the factors affecting the number of victims of suicide in Iraq, based on real monthly data for the period (1/2021-12/2024) issued by the Iraqi Ministry of Interior, using the Invasive Weed Optimization (IWO) and Gravitational Search Algorithm (GSA), as numerical tools to estimate model parameters, the study was conducted in three geographical regions (northern, central, and southern) to reveal spatial differences in the significance of explanatory variables. The results of the IWO algorithm showed that the model has good explanatory power, with an efficiency of 82.5% in the northern region, 85.1% in the central region, and 87.1% in the southern region, with relatively low mean squared error values, indicating acceptable suitability for the model, especially in the southern region. On the other hand, the GSA algorithm showed higher efficiency in estimating model parameters. The study achieved estimation efficiencies of 83.1%, 88.0%, and 88.7% for the northern, central, and southern regions, respectively, with a clear decrease in mean squared error values, particularly in the southern region. The results also showed a consistent effect of some key variables across all regions, while the significance of other variables varied according to geographical specificity. This study confirms the suitability of artificial intelligence algorithms for estimating Poisson regression models for criminal data and their ability to accurately represent spatial variation in suicide cases.