Aplicação de técnicas de agrupamento para avaliar perfis de vulnerabilidade de imigrantes Venezuelanos no Brasil

Abstract

With the significant increase in the flow of Venezuelan immigrants to Brazil in recent years, it is crucial to understand the different profiles within this population to formulate effective public policies and provide appropriate assistance. This study aims to identify these profiles using unsupervised machine learning, specifically the k-means clustering algorithm, utilizing the 2021 dataset provided by the United Nations High Commissioner for Refugees (UNHCR). By applying evaluation metrics such as the silhouette coefficient and the elbow method, it was determined that the optimal number of groups would be two. This segmentation revealed significant differences between immigrant profiles, highlighting greater vulnerability among those who arrived in Brazil more recently, particularly during the COVID-19 pandemic. The results indicate that recent refugees face more intense challenges, likely due to a combination of new economic and social difficulties exacerbated by the global health crisis. These findings underscore the urgent need for targeted public policies that address the specific needs of the most recent and vulnerable immigrants, helping to ensure more effective integration and an appropriate response to the challenges they face. It also highlights the need for a greater understanding of the conditions under which the research was conducted and a concerted effort by the Brazilian government to centralize this information and ensure that immigrants in higher degrees of vulnerability are also heard to avoid potential biases.