Bacterial resistance to antibiotics poses a major threat to global health, intensified by the improper use of antibiotics. The lack of effective and rapid diagnostic tools exacerbates the problem, resulting in less efficient treatments and increased spread of bacterial resistance. Therefore, there is an urgent need to develop models that can assist in more precise and restricted antibiotic prescriptions. Thus, the overall objective of this work is to aid in the construction of binary classification models aimed at determining whether a patient is infected by a given bacterium based on clinical and laboratory data To this end, we utilized data from the Hospital das Clínicas at UFMG. In our previous work, we focused on organizing and cleaning the data to extract relevant patient features and bacterial targets identified through bacterial culture results, enabling the creation of primary datasets that served as inputs and outputs for our models. This process was conducted in collaboration with infectious disease specialists from the hospital through weekly meetings. In this study, we evaluate the performance of three distinct modeling pipelines: a baseline Gradient Boosting-based model optimized using the ROC (Receiver Operating Characteristic) AUC (Area Under the Curve) metric with default parameters, a hyperparameter tuned Gradient Boosting-based model using random search, and a combination of Multivariate Imputation by Chained Equations (MICE) with Z-Score Normalization and a Support Vector Machine (SVM) model. We conducted extensive data preprocessing and filtering, and compared the performance of these approaches using K-fold cross-validation. The hyperparameter tuned Gradient Boosting-based model outperformed the other approaches, achieving satisfactory results given the complexity of the problem.