Stress Analytical Modelling Based on People’s Views on Social Networks

Abstract

With the growing world, the human mind has grown too much with its own complexities. Gone are the days where people used to express themselves through speech or by verbal contact. Now, the era of social media has brought an interface to the world where they can convey their opinions as well as their inner most thoughts through various social networks. People are more comfortable to express their emotions on these social media rather in the real world. This all has led to the need of Sentiment analysis. It has a major role in detecting stress in humans and how surrounding environment is affecting the population of the world. The project analyses the stress among people through tweets. Self-report questionnaires face to face interviews wearable sensors is the main basis of psychological stress that is caused traditionally. The project covers all possible aspects of interactions on social media. Firstly, by fetching tweets from twitter dynamically based on keyword entered by user and segregating them into positive, negative and neutral categories using Naive Bayes algorithm. Secondly, performing sentiment analysis on a dataset containing movie reviews and thirdly, on a very large dataset containing 5 million tweets using Hadoop and an added algorithm of logistic regression for improved performance and efficiency. The entire project was carried out using a distinct step by step procedure consisting of data collection, data cleaning, training of data, data modelling, algorithm application and visualization. Experiments were conducted on an extensive basis to verify the superior theory algorithms and credibility of the project.