UNS Conference Portal, IndoMS International Conference on Mathematics and Its Application (IICMA 2021)

Font Size: 
A comparative study between K-means and K-medoids in the discretization of PGA variables.
Devni Prima Sari, Media Rosha

Last modified: 2021-11-19

Abstract


Indonesia's position on three plates, namely the Eurasian Plate, the Pacific Plate, and the Indo-Australian Plate, makes Indonesia vulnerable to earthquakes. We all know that the science to study the exact time when an earthquake will occur is still limited. So that one of the efforts we can make is to minimize the impact of earthquakes by knowing the level of damage caused by earthquakes in an area. To determine the level of damage, it can be done using the Discrete Bayesian Network. In this Discrete Bayesian Network, we have to make sure all the variables are discrete, including the PGA variables. For this variable discretization process, we can modify clustering algorithms such as K-Means and K-Medoids. In this study, we compared the results of discretization using the two methods. If we use average silhouette widths, we get better discretization using K-Means.