Analisis Sentimen pada Data Ulasan Twitter dengan Long-Short Term Memory

The development of social media has made it easier for people to disseminate information. One form of information in question is the freedom to express opinions on social media. The development of research related to sentiment analysis on text review data aims to determine the polarity of opinion on social media which has increased. One of the methods applied in the sentiment analysis of the review text is the use of the Long Short-Term Memory (LSTM) method. The purpose of this study was to determine the performance of the LSTM model on various sentiments of reviews of Indonesian-language Twitter texts. The testing process is carried out based on the calculation of the hyperparameter tuning accuracy value. Testing the accuracy of this study using Word2Vec parameters, activation function, number of epochs, and number of neurons. The optimal LSTM performance test results from this study were obtained based on tuning the Word2Vec Continuous Bag of Words (CBOW) architecture with an accuracy of 57.15%, tuning the number of neurons as much as 150 producing an accuracy value of 57.35%, tuning the number of epochs at 30 producing an accuracy value of 57. .40%, and tuning the softmax activation function produces an accuracy value of 57.35%.