Nosocomial infections are problems of enormous proportions that impact the well-being of hospitalized patients, as well as the costs associated with treatment. Infection control and management plays an important role in ensuring better quality of life for patients and reducing healthcare costs. Although there are already infection surveillance and prevention programs in place, the reporting of infections is manual and time-consuming, resulting in late detection of this pathology. From this context, this study aims to demonstrate the feasibility of using predictive models for the timely detection of infections resulting from surgical intervention. To do this, we used historical data from the Hospital da Senhora da Oliveira in Guimarães (HSOG) related to various sources. Some ML techniques were used, and two scenarios were developed, with the XGB algorithm standing out as the best model, with recall above 92% and precision above 93%. In the evaluation phase, we also analyzed the business impacts and explored the main variables that have the most impact on predicting infections. The results show that it is possible to develop predictive models capable of predicting the risk of a patient undergoing surgery contracting an infection on a weekly basis.