Topic modeling and sentiment analysis about Mandalika on social media using the latent Dirichlet allocation method

Abstract

The rapid and widespread dissemination of information currently affects the tourism sector. One tourist area that is quite widely discussed is the Mandalika Circuit. Twitter is one platform that provides comments related to the Mandalika Circuit. The amount of information related to the Mandalika Circuit is currently not being utilized properly by managers (government or private). It causes many topics related to the Mandalika Circuit that are currently trending, and public sentiment regarding the Mandalika Circuit is unknown to the government or private sector. Ignorance can result in delays in decision making which can harm the manager. To overcome this problem, research on sentiment analysis and topic modeling related to the Mandalika Circuit was carried out. The sentiment analysis method used is SVM and for modeling using LDA. Based on the results of the sentiment analysis, 1500 tweets were obtained before doing the pre-processing process, thus getting a dataset of 500 tweets divided into 398 positive and 102 negative tweets. So it can be concluded that more Twitter users give positive than negative responses to the Mandalika Circuit. The test results show that the SVM algorithm can classify sentiment toward the Mandalika Circuit well, as indicated by the measurement of the performance of the SVM algorithm, namely 87% accuracy, 77% precision, 84.81% recall, and 98.52% specificity. These results also show that the F1 Score compares the average precision and recall, which is weighted at 80.72%.