Previsão de tempos de internamento num hospital português: Aplicação da metodologia CRISP-DM

Abstract

Using data collected from a Portuguese hospital, within the period 2000 to 2013, we adopted the CRISP-DM methodology to predict inpatient length of stay. The best method (random forest algorithm) achieved a high quality prediction. Such model allowed the identification of the most relevant input features, which are related with the patients' clinical attributes.