Stress Analysis Using Machine Learning

Stress is one of the major health issues of the world and one of the major reasons for committing suicide. Also, it leads to other mental health issues such as depression, anxiety etc., and damage to organs related to respiratory, cardiovascular and nervous systems. In recent years, stress has impacted many individuals due to the pandemic situation. Since the governments across the globe had started to impose lockdowns, the levels of stress significantly raised because of the disturbances led by covid infections, losing loved ones, continuous engagement with laptops and mobiles etc. It is also found that stress has not only disturbed the health condition but also disturbed the relationships and became a self-destruction component. This project is aimed to help those people to understand their stress and consult a psychologist at right time to overcome the situation. Though stress is an active area of research and achieved high performance of models, those were based on signal and speech which were computationally costlier and text-based research work using a state-of-the-art model called the BERT has achieved an f1-score i.e. 80.65%. This project focuses on text-domain and uses open-sourced Stress Analysis on Social Media dataset available on Kaggle which contains 3.6 K samples. In this project, both Machine Learning and Deep Learning Models were trained with 80% of the data and validated with 20% of the data. After, optimization and evaluation of several models, the best model has achieved a benchmark result of 83.74% f1-score on test data using a new network architecture i.e. combination of stacked Transformer Encoder layers with stacked Bi-directional-LSTM. In addition to this, an explainable AI has been implemented for an embedding layer to inspect input attributions in predicting the results. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

A Novel Deep Learning Based Healthcare Model for COVID-19 Pandemic Stress Analysis

Coronavirus (COVID-19) has impacted nearly every person across the globe either in terms of losses of life or as of lockdown. The current coronavirus (COVID-19) pandemic is a rare/special situation where people can express their feelings on Internet-based social networks. Social media is emerging as the biggest platform in recent years where people spend most of their time expressing themselves and their emotions. This research is based on gathering data from Twitter and analyzing the behavior of the people during the COVID-19 lockdown. The research is based on the logic expressed by people in this perspective and emotions for the suffering of COVID-19 and lockdown. In this research, we have used a Long Short-Term Memory (LSTM) network model with Convolutional Neural Network using Keras python deep-learning library to determine whether social media platform users are depressed in terms of positive, negative, or neutral emotional out bust based on their Twitter posts. The results showed that the model has 88.14% accuracy (representation of the correct prediction over the test dataset) after 10 epochs which most tweets showed had neutral polarity. The evaluation shows interesting results in positive (1), negative (-1), and neutral (0) emotions through different visualization. © 2022 Tech Science Press. All rights reserved.

Stress Analysis and Care Prediction System for Online Workers

Working from home (WFH) online during the covid-19 pandemic has caused increased stress level. Online workers/students have been affecting by the crisis according to new researches. Natural response of body, to external and internal stimuli is stress. Even though stress is a natural occurrence, prolonged exposure while working Online to stressors can lead to serious health problems if any action will not be applied to control it. Our research has been conducted deeply to identify the best parameters, which have connection with stress level of online workers. As a result of our research, a desktop application has been created to identify the users stress level in real time. According to the results, our overall system was able to provide outputs with more than 70% accuracy. It will give best predictions to avoid the health problems. Our main goal is to provide best solution for the online workers to have healthy lifestyles. Updates for the users will be provided according to the feedback we will have in the future from the users. Our System will be a most valuable application in the future among online workers. © 2021 IEEE.