The benefits of obtaining a reasonable estimate for the length of stay in Intensive Care Units (ICUs) can be divided into levels, such as assistance, operational and strategic. At the assistance level, a better estimate allows optimizing healthcare protocols (such as sedation and mobilization), a better discussion of the cases in each multi-professional round, and a better preparation of the transfers. At the operational level, it allows better planning of ICU discharge, prioritization of patients to be evaluated daily, and better communication between family members, teams, and managers. At the strategic level, it allows more subsidies for discussion with healthcare operators, a better dimensioning of flow and beds, and a better benchmarking analysis between ICUs. Therefore, this study aims to propose and apply a methodology to predict the length of stay in the ICU based on a data-driven approach and machine learning techniques.