This chapter explores the application of data mining techniques to predict students’ academic performance, providing early identification of underperforming students to enable timely interventions and prevent academic failures. Addressing challenges such as handling diverse and complex student datasets, balancing class distributions, and identifying relevant features, the study employs data preprocessing, classification, and feature selection methods. A minimalistic feature set, SDFS, was developed through experiments on secondary data from the UCI repository, demonstrating superior accuracy compared to other feature sets. Classification models were validated using ten-fold cross-validation, with performance assessed through metrics like classification accuracy and root mean square error. SDFS feature set gives a higher classification accuracy of 91.2% compared with other feature sets. The methodology was applied to both secondary and primary datasets, with the latter designed in consultation with stakeholders. Additionally, the relationship between smartphone usage and academic performance was analyzed, revealing that excessive usage negatively impacts students’ outcomes. A negative correlation of −0.745 was found between academic performance and smartphone usage. This chapter highlights the potential of machine learning and data mining to enhance academic performance prediction and offers valuable insights into its practical implementation in educational settings.