Sentiment Analysis for Preventing HIV Policy Comments Using Word2Vec

Public hearing is important to know how most people feel about the problem they encounter. Proposing policy like HIV prevention is one of them that is currently a social issue. This research presents the experiment of sentiment analysis of social media comments from preventing HIV policy problems, with three classes of the sentiment: positive, neutral, and negative. In addition, to show how the models could get the higher accuracy after fine-tuned parameters. We also used Word2Vec to make word embeddings and combined them with a classifier model and deep learning model. To reach the purpose, we do the grid search to get the best parameter set that gives the high accuracy of each model. The result shows Random Forest achieves the highest accuracy of 95.92%, which is a better performance than other works.