Abstract:
Depression, anxiety, and stress are growing problems worldwide with 3.8% of people
experiencing depression and 4.4% dealing with anxiety. These concerns are
particularly widespread in Pakistan, affecting 20 million individuals in the country.
The treatment of mental illness is hindered by stigma, a scarcity of mental healthcare
resources, prolonged consultations, and the expensive fees associated with consulting
psychologists and psychiatrists. The objective of this study is to develop machine
learning (ML) prediction models for mental illness issues like depression, anxiety and
stress among Pakistani students. The Depression Anxiety Stress Scale 21 (DASS21)
is utilized to gather data from 115 students. A study has introduced a novel process of
data generation considering the utilization of multinomial probability distribution
with correlation. The use of chi-square test and Recursive Feature Elimination (RFE)
with random forest reveals that all 21 features of DASS21 (depression, anxiety, and
stress) exhibit statistical significance. The prediction models are developed using five
machine learning algorithms i.e., random forest, decision tree, support vector
Machines, naive bayes, and k-nearest neighbors. A comparison reveals SVM
outperforms other models with an F1 Score of 0.97, 0.96 and 0.97 for depression,
anxiety and stress. The proposed models could be used in a practical framework to
facilitate the quick screening decision support system for depression, anxiety, and
stress.
