Modelo de predicción de la pobreza en Costa Rica: Una herramienta para los programas de asistencia social

Social assistance programs aimed at combating poverty require a method to define the beneficiaries. The method must be able to identify people in conditions of poverty and even discriminate between those who are in situations of greater poverty or vulnerability since resources are generally limited. One of the most used methods by international organizations and government organizations is the Proxy Mean Test (PMT). This test attempts to predict poverty through an equation that is constructed with a small number of variables from household surveys. Several studies carried out in different countries have shown that the Proxy Mean Test included significant amounts of non-poor households as beneficiaries of social programs and at the same time excluded others that are in poverty (ie Lucia, 2014; Kidd and Wyldi, 2011). There is even a study conducted in Costa Rica (Delgado, 2017), which shows that the PMT may have caused errors of inclusion of 29% in the Advance program and 21% in the IMAS family welfare program. For this reason, in the present investigation, it is proposed to generate a method to predict the degree of household poverty using machine learning algorithms. To train and evaluate the algorithms, the multi-purpose household survey that is carried out every year in Costa Rica by the National Institute of Statistics and Censuses will be used. This research could contribute to a better allocation of resources destined to the attention of poverty, and thus improve the effectiveness of social programs. In addition, it could also be useful to international organizations such as the World Bank that today still use the PMT to identify households in vulnerable conditions

Construir un modelo predictivo del grado de pobreza de los hogares mediante algoritmos de aprendizaje automático, que puede ser utilizado por programas de sistencia social para identificar a los beneficiarios

TICs aplicadas a procesos empresariales, ya que en esta línea se investiga cómo se puede mejorar la toma de decisiones, y “el desempeño en función del cumplimiento de objetivos de organizaciones públicas y privadas, con la implementación y uso de estas tecnologías” (CIADEG, s.f.)