Topic modeling of weather and climate condition on twitter using latent dirichlet allocation (LDA)

Abstract

This study aims to apply topic modeling approach using LDA for Twitter dataset shared by the official Twitter account of BMKG in Java Island. The topic model result can be seen as the representation about what kind of information that posted by BMKG through Twitter. In addition, it can also illustrate the weather, climate, and disaster trends that occurred in Java Island. Based on the topic modeling result, we found five notable topics from BMKG's Twitter accounts. As for the topics discussed, BMKG's Twitter accounts disseminate the information about weather Information and weather forecast; latest weather information in Yogyakarta Region; weather forecast and warning in Central Java and West Java; earthquake information; latest articles and calendar cropping information. In this study, we also illustrated the trends about weather and disaster by analyzing the most frequent words from each region in Java.