Understanding the factors that influence student performance is essential for designing effective strategies to enhance academic success and foster productive learning environments. This study explores the application of Deep Support Vector Machines to predict student performance and identify key variables impacting academic achievement. Unlike prior research, which often focuses on maximizing accuracy, this work prioritizes extracting actionable insights and fostering a deeper understanding of the factors driving student outcomes. Using a comprehensive dataset encompassing academic metrics and engagement data, DSVM models were developed and compared with traditional machine learning approaches, including Random Forest, Gradient Boosting Machine, Logistic Regression, and Neural Networks. While DSVM did not outperform a baseline model utilizing Conditional Generative Adversarial Networks in terms of accuracy, it demonstrated a strong capability for uncovering meaningful patterns in the data. Key determinants of student success were identified, such as exam scores, final grades, adherence to assignment deadlines, and levels of content engagement. This research significantly contributes to educational data mining by emphasizing the extraction of practical and data-driven insights, which can inform targeted interventions and policy-making. Although training DSVM models can be computationally intensive, their ability to highlight critical academic variables positions them as a valuable tool for improving student outcomes. Future work will aim to expand dataset diversity, implement real-time prediction systems, and integrate these predictive models into accessible educational tools. By bridging advanced analytics with actionable outcomes, this study provides a foundation for enhancing personalized learning experiences and promoting student success.