Mental disorders can be recognized by how a person behaves, feels, perceives, or thinks over a period of a lifetime. Nowadays, a large number of people are feeling stressed with the rapid pace of life. Stress and depression may lead to mental disorders. Work pressure, working environment, people we interact, schedule of the day, food habits, etc. are some of the major reasons behind building stress among the people. Thus, stress can be detected through some conventional medical symptoms such as headache, rapid heartbeats, feeling low energy, chest pain, frequent colds, infections, etc. The stress also may reflect in normal behavior while carrying out day-to-day activities. Individuals may share their day-to-day activities and interact with friends on social media. Thus, it may be possible to detect stress through social network data. There are many ways to detect stress levels. Some of the instruments are used to detect stress while there is a medical test to know the stress level. Also, there are apps that analyze the behavior of the person to detect stress. Many researchers had tried to use machine learning techniques including the use of various algorithms such as Decision Tree, Naïve Bayes, Random Forest, etc. which gives a lower accuracy of 70% on average. In this paper, we are using a closeness of stress levels with social media data shared by many users. In our proposed system design, Facebook posts are being accessed using a token. Further, we recommend the use of machine learning algorithms such as Conventional Neural Network (CNN) to extract Facebook posts, Transductive Support Vector Machine (TSVM) to classify posts and K-Nearest Neighbors (KNN) to recommend nearby hospitals. With the help of these algorithms, we predict the stress level of the person as positive, negative. Thus, we are expecting more accuracy to detect the stress along with the preventive recommendation. We have proposed a methodology to detect stress because severe stress may lead to self-harming activities and also it may affect the lives of people around us. Thus, stress detection has become extremely important and we are expecting that our proposed model may detect it with more accuracy.
Personal Stress-Level Clustering and Decision-Level Smoothing to Enhance the Performance of Ambulatory Stress Detection With Smartwatches
