Perceived Stress Prediction among Employees using Machine Learning techniques

Stress is an overworked emotion that can arise from a multitude of factors in our daily lives. It has produced a potentially dangerous scenario for employee mental health throughout the world. To treat and minimize the consequences of stress, there must first be a reliable and reproducible technique for determining the degree of stress being experienced. To date, there have been limited studies to quantify individual stress levels beyond self-reporting by EEG, ECG, questionnaires, and other methods. All of these approaches have an issue with estimating stress accurately. As a result, this inspired me to concentrate on a better understanding of employee stress. In this research, we used the Perceived Stress Scale (PSS) technique for stress prediction. The motivation for using the PSS technique is, an easy-to-use questionnaire with well-established psychometric traits, in which the questions are prioritized using a weighted average approach, and PSS scores are computed by inverting responses, which improves the accuracy even further. In this process, we have collected the data from 251 employees via a questionnaire and used Exploratory Data Analysis (EDA) to visualize the data. According to the results of this study that used the Random Forest, Logistic Regression, and SVM techniques, only about 9.6% of employees are stress-free. From the experimental findings, the logistic regression method gives the highest prediction accuracy of 99 percent.