Predicting student performance has become a crucial task in educational data mining, enabling institutions to identify academic risks early and offer personalized interventions. Machine learning methods are employed in this study to predict student outcomes based on various factors, including demographic information, attendance rates, study hours, previous academic performance, extracurricular activities, and parental support. A synthetic dataset containing information on 1000 students was collected, featuring both quantitative and categorical variables. Data preprocessing involved label encoding, feature scaling, and careful preparation of input features. Several machine learning models were trained and evaluated, including Decision trees, K-Nearest Neighbours (KNN), Random Forest Regressor, Linear Regression, Support Vector Regressor (SVR), and Naive Bayes. Models were evaluated based on their F1 score, precision, and recall, with hyperparameter tuning conducted to optimize performance. The Random Forest Regressor outperformed the other models in terms of prediction, handling intricate, non-linear interactions with ease. The study emphasizes the practical application of machine learning for predicting student success, offering valuable insights for early intervention and academic planning. Although based on a synthetic dataset, the results highlight how machine learning can improve learning outcomes. Future research should focus on using real-world data, exploring advanced algorithms, and improving model generalizability for broader applications.