Predicting Academic Performance: Analysis of Students' Mental Health Condition from Social Media Interactions

Area of Science:

• Computational Social Science
• Educational Psychology
• Natural Language Processing

Background:

• Social media use is pervasive, offering insights into personality, sentiment, and behavior.
• Understanding the link between social media, mental health, and academic outcomes is crucial for student well-being and success.

Purpose of the Study:

• To investigate the association between students' psychological attributes and mental health, derived from social media interactions, and their academic performance.
• To develop and validate a predictive model for academic performance based on social media data.

Main Methods:

• Collected textual data from students' Facebook news feeds using judgmental sampling.
• Utilized MPNet (Masked and Permuted Pre-training for Language Understanding) for feature vector derivation.
• Developed a two-level hybrid model predicting academic performance (GPA) from social media posts via psychological and mental health attributes.

Main Results:

• Identified significant correlations between social media usage, psychological attributes, mental health status, and academic performance.
• The developed hybrid model achieved a high microaverage f-score of 0.94 and an AUC-ROC score of 0.95.
• An ensemble model combining psychological and mental health predictions outperformed independent models.

Conclusions:

• Social media interactions provide valuable data for predicting students' psychological states and academic performance.
• The proposed MPNet-based hybrid model offers a robust and accurate method for assessing student academic potential.
• Further research can leverage these findings to develop targeted interventions for student support.