Optimization Improved K-Means on Centroid Initialization process using Particle Swarm Optimization for Tsunami Prone Area Groupings

Abstract

Tsunami is a high wave caused by tectonic earthquakes, volcanic eruption or landslides in the ocean.  Indonesia is one of the countries that has thousands of islands. Lots of towns is a city on the banks or waterfront city. Indonesia becomes Tsunami prone areas. Tsunami can affect damage in various sectors, namely land degradation and infrastructure, environmental damage, fatalities, even the psychological impact on the victims themselves. Therefore, it takes a clustering of tsunami-prone areas. The result of clustering can give information to the public to remain alert to the danger of the tsunami. Also, clustering of the tsunami can be used by a government to prepare policies in overcoming the danger of the tsunami. Improved K-Means is an approach that proposed in this study to clustering the tsunami prone areas. In selecting the initial centroid must be done properly to produce a high accuracy. We proposed a method to determine the initial centroid appropriately, so that can increase the accuracy. The proposed method is Particle Swam Optimization (PSO). This study also uses comparison methods, such as K-Means, K-Means Improved, and K-Means Improved PSO. This study uses silhouette coefficient to test the accuracy of the system. The result showed that the proposed method has higher accuracy than the comparison method. Silhouette coefficient generated at 0.99924223 with smaller computing time