A HYBRID AI-DRIVEN EDUCATIONAL DATA MINING FRAMEWORK USING FUZZY CLUSTERING FOR ANALYZING SOCIAL MEDIA BEHAVIOR, MENTAL HEALTH, AND ACADEMIC PERFORMANCE

Abstract

This study examines how social media habit and mental health encouragement students’ academic performance. Most traditional methods of evaluating student
performance focus mainly on grades and attendance, often overlooking important factors such as online behavior, stress, and lifestyle habits. To address this gap,
the study considers multiple aspects, including social media usage, levels of stress, anxiety and depression, along with sleep patterns, to better understand their
combined effect on learning outcomes. A hybrid approach is used in this research to handle the complexity of mental health conditions, which are not always
clearly separated. Instead of placing students into fixed categories, the model allows for overlap between different levels of mental health risk. These patterns are
then used along with commonly applied machine learning methods to improve prediction results. In addition, an explainability technique is included to understand
which factors contribute most to the predictions. The study is based on a dataset of 2,500 students. The results show that the proposed approach performs better
than traditional models, achieving higher accuracy. It also highlights that mental health factors, especially stress and anxiety, have a strong negative effect on
academic performance. Social media usage also plays a role, mainly through its impact on mental well-being. Overall, the findings suggest that considering both
behavioral and psychological factors can provide a more realistic understanding of student performance and help institutions identify students who may need early support.