Abstract:
The cardinal feature of social anxiety disorder (SAD) is fear of being judged,
which has an impact on both the emotional and intellectual factors of their
lives. Although most of the Sri Lankan undergraduates are suffering from
SAD, it remains largely undiagnosed. Hence, this research aimed to assess
anxiety levels and develop a prediction model to identify associated factors
of SAD among Sri Lankan university students. A Google form was created
using a validated questionnaire to collect data. Anxiety level was assessed
using the Liebowitz social anxiety scale. Supervised machine learning
methodologies were used to assess the correlations of SAD. Of the total 365
respondents, 47.95% were female, 43.29% were male and 8.77% were not
revealed gender. The majority; 51.51% belonged to the 22–25-year age
group. Among the participants, 78.08% of the students were found to have
some form of SAD, with 38.63% experiencing severe SAD, 23.29%
moderate SAD, and 16.16% mild SAD. The Random Forest Regressor, XG
Boost Regressor, and Linear Regression were tested, and among them, the
Linear Regression model was best fitted with the study results, achieving the
lowest Root Mean Squared Error (RMSE = 2.74 × 10⁻¹⁴), indicating it
performed best in terms of error measurement. According to the model, the
most affected factors for SAD were age (0.305), Year of the study (0.249),
University (0.173), and relationship status (0.132). The study reveals a high
prevalence of SAD among Sri Lankan undergraduates, with significant
correlations to age, study year, University, and relationship status. The Linear
Regression model proved to be the most effective in predicting factors
associated with SAD. These findings emphasize the need for targeted mental
health interventions and support services for students while offering a
foundation for future research in understanding and addressing SAD in this
demographic.
