Clustering of Public Opinion on Natural Disasters in Indonesia Using DBSCAN and K-Medoids Algorithms

Abstract

Natural disasters are disasters caused by events or series of events caused by nature such as earthquakes, tsunamis, volcanic eruptions, floods, tornadoes, and landslides. Some of these natural disasters have taken a lot of public attention, from empathy, sadness and criticism that form an opinion on social media. One of the most popular social media used by the public is Twitter. Opinions written by Twitter users are called tweets. A collection of tweets can be processed to obtain information by using data mining techniques namely Text Mining. In this study, the Density-Based Spatial Clustering of Application with Noise (DBSCAN) algorithm and K-Medoids were used. The result of this study shows that DBSCAN is the best algorithm because it has the Silhouette Index (SI) validity of 0.9140 and the average execution time in RapidMiner Studio is 83.40 seconds. Meanwhile, the K-Medoids algorithm has a Silhouette Index (SI) validity of 0.2259 and an average execution time in RapidMiner Studio 849.93 seconds. The frequency of the word “earthquake” dominates for the positive category, the word “disaster” dominates the negative category, and the word “flood and earthquake” dominates the negative category.