Analysis of Tweets with Emoticons for Sentiment Detection Using Classification Techniques

Social networking services allow users to communicate with their friends and exchange ideas, photos, and videos that delineate their feelings. Sentiments are emotions that express a person's attitude, feelings, and worldview. This raises the possibility of analyzing individual moods and emotions in social network data in order to learn more about people's inclinations and perspectives when communicating online. Sentiment Analysis is the computational study of opinions, assessments, attitudes, subjectivity, and viewpoints represented in text. The emotive appraisal of a condition is a general evaluation of that condition that may be positive or negative depending on physical or mental reactions. In this paper, we attempt to evaluate tweets that contain both text and emoticons in order to determine whether they are positive or negative. This study looked at XGBoost, LinearSVC, Logistic Regression, and BernoulliNB algorithms, with XGBoost providing the highest accuracy of 87.841%. The paper's key contribution is the use of the XGBoost model for tweets that include emoticons, which also produced the greatest accuracy.