Comparative Analysis of K-Means and Random Forest for Student Academic Performance Prediction

The increasing availability of educational data has created opportunities for machine learning (ML) techniques to support academic performance analysis, early identification of students at risk of poor outcomes, and evidence-based educational decision-making. This study presents a comparative analysis of K-Means clustering and Random Forest (RF) for student academic performance prediction. Although both techniques can discover useful patterns in educational datasets, they differ fundamentally in their learning paradigms: K-Means is an unsupervised clustering algorithm that identifies groups of students with similar characteristics, whereas Random Forest is a supervised ensemble-learning algorithm capable of directly predicting predefined academic outcomes. The study adopts the UCI Student Performance dataset, which contains demographic, social, school-related, behavioural, and academic variables. The proposed methodology involves data preprocessing, feature transformation, exploratory analysis, K-Means clustering, Random Forest classification, and comparative evaluation using accuracy, precision, recall, F1-score, cluster quality, and computational considerations. Recent studies indicate that tree-based methods, particularly Random Forest and related ensemble models, remain widely used in student-performance prediction, while clustering provides useful insights into heterogeneous student profiles. The analysis demonstrates that the two algorithms should not be regarded as direct substitutes: K-Means is primarily valuable for discovering latent student groups, whereas Random Forest is more appropriate when the objective is explicit prediction of academic-performance categories. The study therefore highlights the complementary rather than identical roles of clustering and supervised learning in educational data mining.