Type 2 diabetes a primary contributor to global mortality rates, with a significant impact in Metropolitan Lima, especially among people over 30 years of age. As technological tools advance, the application of predictive modeling through Machine Learning (ML) has grown crucial for enhancing the early identification of this condition. This article presents intelligent models based on supervised learning methods, including Decision Tree (DT), Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM). The study centers on creating and assessing predictive models via a meticulous data preprocessing workflow, which incorporates a strategic blend of SMOTE with LOF for data balancing and employs cross-validation to guarantee model precision and resilience. The results highlight the KNN model as the most effective, achieving superior performance after iterative tuning and dataset optimization. This study contributes to diabetes research by implementing advanced preprocessing and model selection methodologies, enabling more accurate and reliable detection, thus helping to reduce the medical and financial burden associated with this disease.