Sentiment Analysis Terhadap Tulisan Mengenai Universitas Pertamina Di Media Sosial Twitter

Abstract

Nowadays, branding is not only applied to business sectors. Many universities have implemented branding to maintain the university's brand image so that it can attract prospective students. Sentiment analysis is one of the activities to monitor the university’s brand image in the community. This research applies a machine learning approach to conduct a social media analysis of Twitter. The dataset consists of 4323 tweets containing the word ”Universitas Pertamina”. Each tweet is then grouped into three different classes based on the tweet's sentiment, namely negative, neutral and positive. This study uses three types of machine learning models: Convolutional Neural Network (CNN), Long Short Term Memory (LSTM) and CNN-LSTM. The procedure for the three models uses the stratified 5-fold cross-validation technique. Model performance is then measured using a learning curve and measuring parameters such as accuracy, weighted-f1 and balanced accuracy. The CNN-LSTM model records the highest accuracy value, 76.92% and the highest weighted-f1 with 76.78%. Meanwhile, the LSTM model gets the highest balanced accuracy value with 66.05%.