Clasificación laboral en México usando un enfoque de aprendizaje automático

In this study, work discouragement in Mexico is addressed from a mathematical modeling perspective. Two conditions of employability are considered: unemployed and discouraged, and the classification of these groups is characterized using machine learning models and sociodemographic variables, such as educational level, sex, age, marital status, number of children, relationship, and area of residence. Considering data from the National Occupation and Employment Survey, the highest classification accuracy of the algorithms addressed was obtained by neural networks and random forests. These models indicated that the main features that distinguish the discouraged from the unemployed are women aged 20-29, with high school and higher education, without children, single, and residing in urban areas. The most relevant thing is that, thanks to the results obtained with the machine learning models, it is possible not only to predict with greater precision who could fall into work discouragement, but also, to propose more effective and focused public policies. These policies can be specifically aimed at the sectors identified as most vulnerable, thus contributing to the reduction of job discouragement and the improvement of employability in the country.