Date of Award
Doctor of Philosophy (PhD)
Tweets are usually the outcome of peoples’ feelings on various topics. Twitter allows users to post casual and emotional thoughts to share in real-time. Around 20% of U.S. adults use Twitter. Using the word-frequency and singular value decomposition methods, we identified the behavior of individuals through their tweets. We graded depressive and anti-depressive keywords using the tweet time-series, time-window, and time-stamp methods. We have collected around four million tweets since 2018. A parameter (Depressive Index) is computed using the F1 score and Mathews correlation coefficient (MCC) to indicate the depressive level. A framework showing the Depressive Index and the Happiness Index is prepared with the time, location, and keywords and delivers F1 Score, MCC, and CI values.
COVID-19 changed the routines of most peoples' lives and affected mental health. We studied the tweets and compared them with the COVID-19 growth. The Happiness Index from our work and World Happiness Report for Georgia, New York, and Sri Lanka is compared. An interactive framework is prepared to analyze the tweets, depict the happiness index, and compare it. Bad words in tweets are analyzed, and a map showing the Happiness Index is computed for all the US states and was compared with WalletHub data. We add tweets continuously and a framework delivering an atlas of maps based on the Happiness Index and make these maps available for further study.
We forecasted tweets with real-time data. Our results of tweets and COVID-19 reports (WHO) are in a similar pattern. A new moving average method was presented; this unique process gave perfect results at peaks of the function and improved the error percentage.
An interactive GUI portal computes the Happiness Index, depression index, feel-good- factors, prediction of the keywords, and prepares a Happiness Index map. We plan to create a public web portal to facilitate users to get these results. Upon completing the proposed GUI application, the users can get the Happiness Index, Depression Index values, Happiness map, and prediction of keywords of the desired dates and geographical locations instantaneously.
Sadasivuni, SUDHA TUSHARA, "Analyzing Tweets For Predicting Mental Health States Using Data Mining And Machine Learning Algorithms." Dissertation, Georgia State University, 2021.
File Upload Confirmation