Dhifan Diandra Heriswan, Yuita Arum Sari, Muhammad Furqon
Twitter is a microblog-based social media platform that can upload text, images, and videos with a limit of 280 characters per tweet. There are various information and opinions obtained on this platform; one of the topics discussed in quarantine during the Covid-19 pandemic. Based on the information and opinion, there are various kinds of responses. The kind of responses could be either pros or cons toward the topic. Sentiment analysis is one way to determine the sentiment response, whether positive, negative, or neutral. There are various ways to apply sentiment analysis, one of which is using unsupervised or clustering techniques. The processes carried out in this study included preprocessing, feature vector extraction, hyper-rectangle, DENCLUE, and K-Means. Preprocessing, feature vector extraction and hyper-rectangle are used for adjusting input values in the clustering process, such as converting data into structured, normalizing, and reducing dimensions. The DENCLUE method is used to remove noise, which can complement the K-Means method because the clustering process does not have noise detection. In addition, there is a sentiment expression dictionary that is used to determine the initial value in clustering calculations. In this study, we conducted tests with the two conditions of sentiment expression (using neutral sentiment expressions and without using neutral sentiment expressions) and various group divisions. Based on the analysis and testing results, the optimal value was obtained in the conditions using a neutral sentiment expression dictionary with a value of k = 2. The value of the resulting silhouette coefficient is 0.6218 with a proper or appropriate cluster interpretation (medium structure). © 2021 ACM.
University of Brawijaya, Indonesia; Faculty of Computer Science, Brawijaya University, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock