| Autor: |
Tirumanadham, N S Koti Mani Kumar, S, Thaiyalnayaki, M, Sriram |
| Zdroj: |
International Journal of Information Technology; December 2024, Vol. 16 Issue: 8 p5429-5456, 28p |
| Abstrakt: |
The paper presents a new feature selection technique developed in detail here to address improved prediction accuracy not only for the machine-learning algorithm but also for student outcomes related to the learning environment. It is a feature selection methodology, combined with Ridge (L2) regularization and the Boruta optimization clubbed in a two-tier approach named as BR2-2 T, which results in the most effective feature set selection process. To be specific, the present study developed a three-tier ensemble model that consolidates Random Forest with Bayesian Optimization methods Support Vector Machine with random search, and Gradient Boosting with Particle Swam Optimization (PSO) for performing hyperparameter tuning. Some sophisticated techniques applied in this study are the use of Z-score normalization so as to standardize variable distribution, Synthetic Minority Over-sampling Technique (SMOTE) for handling imbalanced dataset and Multiple Imputation by Chained Equations (MICE) for handling messiness among items. The evaluation of the projected approach shows improvement in accuracies, which can attain a maximum of 98.74%. The comparative assessment of the proposed strategy unveiled that predictive modelling of educational outcomes is effectively enhanced more than common methods of topics related to education. The research sharpens student success strategies and educational pedagogy by using powerful algorithms in its direction of decision-making. |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
Full text from SpringerLink