Hybrid Data Mining Techniques for Predicting Student Academic Performance in E-Learning to Avoid Drop-Out (HDL-SP): An Efficient Data Mining Technique to Forecast Academic Performances of Students

Hybrid Data Mining Techniques for Predicting Student Academic Performance in E-Learning to Avoid Drop-Out (HDL-SP): An Efficient Data Mining Technique to Forecast Academic Performances of Students

연구 분야: Databases

학회: SN Computer Science

It is essential to students' performance in learning organizations, which are widely used as gauges for activities and academic establishments. Preventive interventions can significantly boost the success of susceptible learners; hence the primary objective should be to identify them early. Though a lot of research has been done to date to address this problem, state-of-the-art methods are still unable to greatly improve performance estimate, which raises the likelihood that learners will be sensitive. A hybrid data mining technique, named HDL-SP, is proposed to predict student academic performance in e-learning environments and prevent dropouts. The approach uses a Tunicate Swarm Grey Wolf Optimization-based clustering method for data preprocessing to remove redundant information. To reduce dimensionality, a Teacher Learning-based Reactive Search Optimization (TL-RSO) algorithm selects optimal features from an educational dataset. Performance prediction and classification are achieved using hybrid reverse transfer learning-based deep belief networks. The effectiveness of HDL-SP is evaluated using publicly available datasets. Compared to existing techniques such as Random Forest, Neural Networks, and KNN, HDL-SP achieves substantial improvements across various metrics, including a 23% average increase in accuracy, a 12.9% average increase in precision, a 22.2% average increase in recall, and a 19.89% average increase in F-Measure. Furthermore, the proposed scheme has been benchmarked against baseline models, with the results consistently highlighting HDL-SP's superior performance in predicting student academic success and reducing dropout rates in e-learning environments.