Study of victimization and perception of insecurity in Ecuador: application of hybrid models with logistic regression, machine learning, and Neutrosophy

Abstract

Given the victim's rising profile in contemporary victimology and criminology, this study examines victimization and the impact of perceived insecurity, paying particular attention to the situation in Ecuador, where reports of crime and violence in urban areas are on the rise. The overarching goal is to create a reliable model for victimization risk assessment that may also serve as an analytical tool for public policymaking and preventative efforts. One step is to build a classic logistic regression model that takes into account variables like age, sex, education level, occupational category, and level of physical activity. The goal is to understand the factors that increase or decrease the likelihood of being a victim of a crime in this city or another. By analyzing the coefficients, we can determine the statistical weight of each factor in this victimization phenomenon. Second, we add a machine learning model that can capture complex and non-linear relationships between variables (Random Forest) to improve accuracy over traditional methods and strengthen the predictive capacity. The theoretical-methodological framework of neutrosophic statistics is utilized to convert the variables in victimization surveys, which contain uncertainty and indeterminate values such as "don't know" or "no information" responses, into truth (T), indeterminacy (I), and falsity (F). This allows for the explicit management of uncertainty and strengthens the validity of the predictions. A fresh approach to victimization research, this synthesis of Neutrosophy, machine learning, and logistic regression provides a scientifically sound framework for comprehending and reducing citizen insecurity.