SPATIAL NON HIERARCHICAL CLUSTER ANALYSIS WITH K-MEANS++ AND SEMIVARIOGRAM MODELING CASE STUDY: EARTHQUAKE DATA IN EAST INDONESIA

• Main Author: Lutfi, Lusiana
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/38948
• Institution: Institut Teknologi Bandung

The Eastern Indonesian area is very vulnerable to earthquakes. The earthquakes that occured have various magnitudes and depths. The earthquakes can be grouped into several clusters based on its location, magnitude, and depth. K-Means++ cluster analysis can be applied to this case to get the more stabilized results. K-Means++ is an alternative to overcome the weakness of k-means that is to minimize the risk of changing the cluster results as the consequence of the random centroid selection. Then, the spatial correlation of the earthquakes’s location in a cluster can be discovered through Semivariogram Modeling. Sill ???? parameter is estimated by variance of each cluster and range ???? is estimated by practical simulation by entering some values between minimum distance and maximum distance until it gives the minimum value of SSE. The best fitted semivariogram model also chosen based on the minimum SSE. For the case of earthquake in 2008, the Gaussian model fit the experimental Semivariogram based on data obtained. The estimated value of ???? always increase from first cluster to third cluster. It means that more dangerous the earthquake, the wider its zone of influence. By considering some highlighted cities and regencies in Eastern Indonesia to the model, it is found that several cities like Ambon, Ternate, and Tidore should be aware of earthquake danger.