<p>Cardiovascular disease (CVD) remains the leading cause of death worldwide, highlighting the urgent need for early and accurate risk prediction tools to reduce mortality and improve patient outcomes. This research uses the Framingham Heart Study dataset to assess how well different machine learning models predict the 10-year risk of CVD. Support Vector Classifier (SVC), Decision Tree, Random Forest, K-Nearest Neighbors (KNN), and a Deep Learning neural network are among the models evaluated. Accuracy, precision, recall, and F1-score were used to gauge the model’s performance. At 91%, the Random Forest Classifier was the most accurate, followed by SVC and Decision Tree at 90%, Deep Learning at 90%, and KNN at 89%. Class imbalance is a problem that is common in clinical datasets, as evidenced by the fact that all models did well in predicting the majority class (low-risk), but significantly worse in identifying the minority class (high-risk). The study emphasizes the significance of cost-sensitive learning, resampling strategies like Synthetic Minority Over-sampling Technique (SMOTE), and sophisticated feature engineering to improve recall for the minority class in order to counteract this. In order to improve clinical relevance and predictive accuracy, future research will investigate ensemble approaches and model interpretability. The results highlight the necessity of striking a balance between overall accuracy and trustworthy high-risk patient identification in order to provide clinical practice with efficient cardiovascular risk prediction and decision support.</p>