The early prediction of students’ academic performance represents a pivotal area of focus within the field of educational research, with substantial implications for enhancing educational outcomes and interventions. This paper offers a comprehensive overview of the current methodologies employed in educational data mining (EDM) for the prediction of students’ academic performance, utilizing a range of data analytics techniques. The study examines significant research utilizing machine learning algorithms, with a particular focus on predictive modeling, clustering, and classification. Furthermore, the paper puts forward a novel hybrid approach that integrates clustering techniques with classification algorithms, to enhance the accuracy and robustness of performance predictions. The methodology comprises two phases: training and prediction. The training phase employs a combination of K-means and hierarchical clustering, while the prediction phase utilizes classification algorithms such as Naive Bayes, Random Forest, and Decision Trees. The experimental results, utilizing a dataset of 1000 students with diverse demographic, academic, and behavioral features, demonstrate the effectiveness of the proposed hybrid model, achieving superior accuracy in predicting academic outcomes. This study underscores the potential of hybrid approaches in addressing challenges such as the cold start problem and their utility in the educational domain for more precise and actionable insights into student performance.