ABSTRACT
The prediction of student performance has become imperative in education, prompting the need for advanced methodologies to anticipate academic outcomes accurately. Currently, there is no most precise way to forecast every variable related to student performance. Extreme Learning Machines are being used in this work to forecast student performance, which will help in the development of tools for higher academic achievement and individualized learning pathways. Extreme Learning Machines are the best choice for early preferences and customized support networks because of their ease of use and capacity to manage sizable database systems. The study of the prediction model-building process includes feature engineering, preprocessing, and data collection. The following algorithms are compared with respect to training times and accuracy: Support Vector Machines, Ada boost, XG Boost, Decision Trees, Random Forests, and Perceptrons (back-propagation). Results indicate that ELMs perform better than other algorithms with 99.84% accuracy and 6.84 ms training time. Furthermore, according to feature importance analysis, the three factors that have the biggest effects on academic achievement are study time, social media use, and parental participation. Teachers and other stakeholders will obtain important insights into students’ progress by putting the ELMs into practice, this allows for the prompt provision of interventions and support systems. By demonstrating the effectiveness of ELMs in forecasting student progress and promoting a culture of academic excellence and equity, the study advances educational analytics. Overall, this research highlights how predictive modeling with ELMs may revolutionize education and improve student success through evidence-based decision-making and targeted interventions.
