On Developing Generic Models for Predicting Student Outcomes in Educational Data Mining

Poor academic performance of students is a concern in the educational sector, especially if it leads to being unable meet minimum course requirements. However, with timely prediction students’ performance, educators can detect at-risk students, thereby enabling early interventions for supporting these overcoming their learning difficulties. majority studies have taken approach developing individual models that target single while models. These are tailored specific attributes each amongst very diverse set possibilities. While this yield accurate some instances, strategy associated limitations. In many cases, overfitting take place when data small or new courses devised. Additionally, maintaining large suite per significant overhead. This issue be tackled by generic and course-agnostic predictive model captures more abstract patterns able operate across all courses, irrespective differences. study demonstrates how developed identifies wide variety courses. Experiments were conducted using range algorithms, producing an effective accuracy. The findings showed CatBoost algorithm performed best on our dataset F-measure, ROC (receiver operating characteristic) curve AUC scores; therefore, excellent candidate providing solutions domain given its capabilities seamlessly handle categorical missing data, which frequently feature datasets.