Data analytics is emerging as a critical field to intelligently utilize the vast trail of data we create in our daily lives. An analysis of data trends can reveal patterns that can predict human behavior in areas such as health care, Ecommerce and consumerism, among others. The purpose of this experiment was to study the correlation between a Twitter hashtag’s sentiment and its trending duration using IBM Watson Analytics. The hypothesis was that a major event associated with a more positive sentiment would trend longer than more negatively associated counterparts. The experiment relates to Hedonic adaptation, the psychological theory that states that humans will return to a relatively happy state despite a negative or positive turn of events. The sentiment was first analyzed on a smaller scale by randomly selecting 30 tweets within each hashtag studied and then on a larger scale using IBM Watson Analytics. For the trend analysis test, the total number of tweets for each hashtag was recorded daily. Manual sentiment analysis yielded a strong correlation of “happy” sentiment with entertainment hashtags, “sad” with natural disaster, “fearful” with health and medicine, and “neutral” with the control group #selfie. A Chi Square Test for Independence was run at alpha = 0.05 on the average number of tweets for the hashtags in each category and showed a direct correlation between the category and sentiment X2 (15, N = 120) = 37.731, p<0.05. Thus, the hypothesis was supported because the entertainment hashtags with positively associated sentiments trended longer than more serious hashtags exhibiting negative sentiments, and there was a direct correlation between the category of the tweet and its sentiment.
"The Effect of Trending World Events on Sentiment Analysis and Relevance Intervals Using Data Analytics Software on Twitter Data,"
Journal of the South Carolina Academy of Science: Vol. 16
, Article 8.
Available at: https://scholarcommons.sc.edu/jscas/vol16/iss2/8