Using Data Mining to Predict Secondary School Student Performance for Zambia

Predicting student performance remains a challenge in many education systems, especially in developing countries like Zambia, where robust predictive tools are scarce. This study shows how data mining methods can be utilised to improve the accuracy of performance prediction by leveraging mock examination results to meet the needs of school management. For this study, a dataset containing 1,170 instances and 17 attributes was constructed and analysed using four classification algorithms (J48, PART, BayesNet, and Random Forest). The findings indicate that although each classifier produced results with high accuracy above 99 percent, Random Forest performed best, delivering perfect predictions with 100 percent accuracy. These results emphasise the importance of data mining in generating reliable forecasts of student performance, enabling early detection of at-risk learners and timely interventions by school managers, teachers, and parents. The study recommends adopting Random Forest as the most suitable classifier for predicting student performance. By incorporating predictive analytics into educational management, schools can strengthen decision-making, refine teaching approaches, and ultimately improve learning quality.