Sentiment Analysis of Presidential Candidate Debates from YouTube Videos

The upcoming Indonesian presidential election holds immense democratic significance. Candidate debates, hosted by prominent journalist Najwa Shihab on her YouTube channel, play a crucial role in articulating visions and addressing national concerns. These debates are pivotal in amplifying public discourse and serve as primary information sources for the electorate. This research presents an extensive evaluation of various machine learning models for sentiment analysis, focusing on their performance metrics in identifying positive sentiments within Presidential Candidate Debates from YouTube videos. Models such as Complement Nave Bayes, Multinomial Nave Bayes, Bernoulli Nave Bayes, Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) were scrutinized. Notable highlights include Bernoulli Nave Bayes and LSTM exhibiting exceptional precision rates of 99.85% and 100%, respectively, showcasing their proficiency in accurately identifying positive sentiment instances. However, concerns of potential overfitting due to these high precision scores were raised, prompting the need for validation across diverse datasets to ensure generalizability. The findings underscore the effectiveness of these models in sentiment analysis while emphasizing the importance of further assessment for broader applicability beyond the specific dataset used in this analysis.