The Influence of Social Media on Stress Detection Using TF-IDF and Machine Learning Strategies

Stress is an intrinsic aspect of human existence, frequently provoked by a range of factors such as work responsi-bilities and social relationships. Although short-term stress might have beneficial effects, long-term stress can result in adverse health outcomes. In modern times, social media has become a venue for individuals to openly convey their emotions, especially feelings of stress. This study investigates the correlation between the utilization of social media and the identification of stress, employing advanced machine learning methods. We utilize the TF-IDF method to extract features and assess several machine learning techniques on refined models. The contribution on this research is applying TF -IDF and several machine learning algorithms on stress detection. The results showed Logistic regression algorithms provide exceptional performance with the accuracy 0,9 and 0,79 respectively in both Reddit and Twitter datasets, exhibiting high accuracy, precision, and recall in accurately recognizing occurrences of stress. The Support Vector Machine (SVM) exhibits potential, especially in identifying stress-positive cases, albeit with significantly poorer precision when compared to logistic regression and random forest. These findings highlight the capacity of machine learning methods to comprehend the influence of social media on the expression of stress.