A Computational Sentiment Analysis of Public Opinion on Nigerian 2023 Presidential Candidates from YouTube Video Comments

YouTube has become an important platform for public expression, where users post comments to communicate their views and feelings about political issues and candidates. This study examined public sentiment and prominent words associated with presidential aspirants in Nigeria's 2023 presidential election using YouTube comments. Comments available online for 16 presidential aspirants who declared their intention to contest the 2023 presidential election were collected using the Google YouTube Data API. Sentiment analysis and word-cloud analysis were performed to identify the dominant emotional patterns and frequently occurring terms in the comments. In addition, the sentiment classifier was evaluated against a gold-standard annotated dataset using accuracy, precision, recall (sensitivity), F1-score, specificity, negative predictive value, balanced accuracy, and Cohen's kappa to assess its classification performance and agreement beyond chance. The results showed that negative sentiment constituted the largest proportion of the comments, followed by positive sentiment, whereas surprise had the lowest proportion. The most frequently occurring word was “Nigeria,” while the names “Tinubu,” “Sowore,” and “Orji Kalu” were prominent in the word cloud. The classifier achieved an accuracy of 78.39%, with a positive-class precision of 62.94%, recall of 93.04%, F1-score of 75.09%, specificity of 70.49%, negative predictive value of 94.95%, balanced accuracy of 81.77%, and Cohen's kappa of 0.5722. These findings demonstrate the sentiment patterns expressed in YouTube comments surrounding the 2023 Nigerian presidential election and provide an empirical assessment of the reliability of the sentiment classification approach.