Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- 4ac16d66-6647-420f-a8d6-e6d68c0b6f53
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted an activity on enhancing machine learning algorithms through recursive feature elimination for student performance prediction. The activity involved researchers from UDINUS: Wijiyanto, Aris Marjuni, Ahmad Zainul Fanani, and Ruri Suko Basuki. The activity addressed prediction systems for student performance in higher education institutions and their stated importance for improving educational outcomes.
The activity considered big data obtained from various educational channels, including examination centers, academic systems, registration departments, and e-learning platforms. The source describes the extraction of meaningful insights from educational data as challenging. In response to this challenge, the activity applied data dimensionality reduction. UDINUS introduced an enhanced recursive feature elimination algorithm as a wrapper feature selection technique for the described prediction task.
The proposed recursive feature elimination approach was combined with five classifiers. These were the Gradient Boosting Classifier, Random Forest Classifier, Support Vector Classifier, K-Neighbors Classifier, and Decision Tree Classifier. The dataset was originally intended to influence factors in dropout students. It contained 329 records and 9 features. The described methodology was tested with the Decision Tree Classifier, and the AUC score improved by 8% compared with the model tested without the proposed feature selection method.
The activity relates to the stated educational goal through its focus on student performance prediction, educational data, and higher education institutions. The source states that educators and academic institutions can utilize the resulting information to develop targeted intervention strategies, tailored learning experiences, and personalized recommendations for students. These uses are presented in the source as ways to promote a more effective learning environment and improve overall educational outcomes. Within the documented scope, UDINUS presents machine learning, recursive feature elimination, classifier testing, and educational data analysis as components of the activity.