On Early Prediction of Risks in Academic Performance for Students Based on Hybrid Gbdtrbm Model

Better learning outcomes and academic performance are the results of educational data mining's ability to predict dangers for students early, allowing interventions to happen sooner. Students can use accurate success projections to guide their course selection and study habits. They also aid educators in identifying struggling kids, which reduces the need for disciplinary actions like expulsions and formal warnings. Methods in the proposed approach include preprocessing, feature extraction, and training the model. The goal of preprocessing is to guarantee high-quality inputs by cleaning and segmenting data. In order to improve the accuracy of factor estimate, feature extraction is used to decrease the number of data dimensions. To improve the accuracy of the predictions, they trained the models using hybrid GBDT-RBM models, which include the best features of both GBDT and RBM. Compared to standard GBDT and RBM models, this one outperforms them with a 95.12 % accuracy rate. The results indicate that sophisticated hybrid modelling has the potential to significantly improve academic performance risk prediction in its early stages. This allows educators to provide individualized support to pupils, which in turn boosts their engagement, retention, and performance in the long run.