Predicting Victimizing Events in the Colombian Armed Conflict from Sociodemographic Variables Using a Neural Network Model

Introduction: Colombia has endured over 60 years of armed conflict. Objective: This study addressed the issue of victims of the Colombian armed conflict through data analysisutilizing Deep Learning (DL) tools. Method: The dataset used is Colombia’s Registro Único de Víctimas (n = 173,994 records), which includes various types of violent events. A neural network model was employed to analyze and predict patterns observed in these situations, focusing on aspects such as the type of event, ethnicity, gender, and disability status. Results: The model achieved an overall accuracy of 79.38% on the test set. The results reveal how these variables are distributed among victims, providing a deeper understanding of the unique characteristics of each group. Discussion: The results visualize the magnitude of the impact on each population group and show how factors such as ethnicity, sex, disability status, and life stage are related to victimization in the Colombian armed conflict.Conclusion: These findings have significant implications for the formulation of policies and support actions aimed at mitigating the impact of the conflict on vulnerable populations.