Implementasi Aplikasi Sentimen Pada Data Twitter Jelang Pemilu 2024

Elections are one of the most important democratic processes, giving citizens the right to choose their leaders. In today's digital era, social media is an increasingly important information source influencing public perception. Twitter has been a social media from the past until now that still exists in finding information. Tweets are one of the most frequently used services to express opinions or opinions to the public. Sentiment analysis as an application of Natural Language Processing (NLP) is helpful in understanding public opinion towards prospective leaders and issues discussed during election campaigns. The motivation for this study is to conduct text classification using a deep learning model called LSTM and to compare the use of oversampling and non-oversampling methods. This research started by collecting datasets from Twitter, labelling, pre-processing, creating and evaluating the model, and implementing it into the web application. The experiment showed that the random oversampling technique gets more significant accuracy than non-oversampling. Random oversampling produces an accuracy of 0.82 at epoch 25, while non-oversampling reaches an accuracy of 0.61 at epoch 50