Ensemble Method for Predicting Student Performance and Dropout Risk

Failing to produce the desired output from students can have negative impacts not only on the students but also on the institutions. Addressing these issues, this paper suggests an ensemble machine learning model to predict student performance and dropout risk. The goal is to build a model that is capable of achieving effective results in predicting student performance, regardless of the characteristics of the dataset. The ensemble method is an effective technique where multiple machine learning models are integrated for a more reliable result. There are several approaches followed in the ensemble methods, such as voting, stacking, and others. In this model, stacking was implemented. Several classification algorithms, such as random forest, logistic regression, support vector machine, and CatBoost were utilized to build the ensemble model. To evaluate the robustness and reliability of the model, three diverse student performance datasets were used. These datasets consider academic records, sociodemographic information, family background details, students’ behavioral information, and others. To assess the model, three evaluation metrics were employed: accuracy, precision, and recall. In all three datasets, the model succeeded in producing the best results. The highest accuracies achieved in the three datasets are 91.59, 89.47, and 88.54%.