An SVM-Based Approach to Predicting Level of Job Anxiety in Corporate Professionals using Linguistic Markers on Twitter

Unnathi Utpal Kumar

Abstract: Work anxiety is linked with decreased job commitment and satisfaction. Yet, work anxiety, like other mental health problems, is not physically diagnosable. The lack of diagnosis and cure of job anxiety leads to lower levels of economic productivity and adds to the mental health epidemic. This study proposes a machine learning model to predict a corporate professional’s level of work anxiety using their tweets by identifying linguistic markers associated with work anxiety. The Twitter API was used to create a dataset of over 15,000 corporate professionals. Thousands of tweets were collected from these users over 3 periods of time (May–August 2019, May–August 2020, and January-April 2021). After conducting sentiment and linguistic analysis, tweets from 90 random users (manually labeled for their job anxiety scores according to the job anxiety scale) were used to train/test an SVM regression model. The model achieved an RMSE of 0.2 and an accuracy of 83%. This approach has the potential to enable early detection of work anxiety and alert individuals about their mental health.