Robust kernel extreme learning machines for postgraduate learning performance prediction

Area of Science:

• Educational Psychology
• Machine Learning Applications
• Higher Education Studies

Background:

• The mentor-mentee relationship is crucial in Chinese graduate education, but recent issues have impacted learning quality.
• Understanding the interplay between mentor relationships, student motivation, and academic pressure is vital for improving graduate outcomes.

Purpose of the Study:

• To investigate how the mentor-mentee relationship influences postgraduate learning performance, moderated by admission motivation and learning pressure.
• To develop and validate a novel machine learning model robust to data outliers for predicting postgraduate academic success.

Main Methods:

• A robust kernel extreme learning machine (RK-ELM) model was developed to handle data outliers and enhance prediction accuracy.
• Data was collected from 873 full-time postgraduate students in Zhejiang Province, China, including questionnaire results and GPAs.
• The study analyzed the predictive power of mentor relationships, admission motivation, and learning pressure on academic performance.

Main Results:

• The RK-ELM model demonstrated effectiveness in predicting postgraduate learning performance.
• The mentor-mentee relationship significantly influences learning performance, but not in isolation; it acts indirectly through learning pressure.
• The combined effect of mentor relationships and enrollment motivation effectively predicts learning performance.

Conclusions:

• The mentor-mentee relationship is a key factor in postgraduate success, mediated by student motivation and pressure.
• The proposed RK-ELM model offers a robust approach for analyzing complex educational data and predicting student outcomes.
• Interventions focusing on strengthening mentor relationships and managing student pressure are recommended to enhance graduate education quality.