Velho, IolandaAlfaro, Miguel Alexandre Rocha Martins2023-06-162023-06-162023-01-092023-01-09http://hdl.handle.net/10362/154017The overcrowding phenomenon is a worldwide problem that has been negatively affecting both public and private hospitals. A suitable and efficient planning of ED resources may diminish the effects of this event. Therefore, a Linear Regression, SARIMAX and Long-Short Term Memory models were developed to forecast weekly ED arrivals. Based on a Machine Learning multi-step ahead predictive tool to help in the decision-making process, the hospital may ensure a good quality of services. First, the predictive tool was used to forecast weekly ED demand for all patients in a big unit of a private Portuguese healthcare provider, CUF, and then, to predict the Urgent Patients weekly ED arrivals for the same unit.engHealthcareEmergency departmentMachine learningTime seriesMulti-step-ahead forecastingmaster thesis203310586