QUANTITATIVE ASSESSMENT AND CLASSIFICATION OF PERCEIVED STRESS -A STATISTICAL AND MACHINE LEARNING APPROACH USING PSS-10 SCALE

Abstract

This thesis outlines stress factors that affect military personnel’s and student’s performance in various situations. The present thesis comprises two objectives, Objective 1: identified the different factors in military context that contribute to military suicide, taking into account psychological, personal, social, behavioral and administrative issues. Under this author, we identify the key factors leading to the surprisingly high suicide rate among soldiers. Gaining understanding of these elements is essential to improve the welfare, security and efficiency of military personnel. The study compares the results and findings of all selected studies to identify the most crucial factors. It was also found that there is no significant difference in stress levels between the different ranks of officers and soldiers in the military. Under Objective 2: We examine the performance of machine learning algorithms which help in early prediction of stress among university students using the best prediction model. The dataset was taken from postgraduate (Master of Technology) students using Google form , it consisted of 57 students data. We have applied 6 types of classification algorithms: Logistic (75.00%), KNN (83.33%), SVM (66.67%), the Decision Tree (92.00%), Random Forest (92.00%)and Naive Bayes’ (91.67%) and also we calculated their accuracy with the help of confusion matrix. In this study, Decision Tree algorithm and Random Forest algorithm shared an equal and highest accuracy of 92.00% as compared to other algorithms. Objective 3: Construct validity of the perceived stress scale (PSS-10) in a sample of university student’s dataset. The result of our study showed that the PSS-10, as a two factor model with good reliability and validity, may accurately measure the stress levels among university students.