INISIALISASI PUSAT KLASTER (CENTROID) AWAL PADA K-MEANS DENGAN METODE CENTRONIT; CENTROID INITIALIZATION FOR K-MEANS USING CENTRONIT METHOD

SILVIA, MARIA, Abdurakhman

2016 | Skripsi | FMIPA

K-Means is one of very well known clustering algorithm because its ability to classify large data very quickly. However, clustering performance of the K-Means highly depends on the initial centroids. Usually initial centroids for the K-Means clustering are determined randomly. Difference in the initial centroid will make a difference in the results of clustering. If the initial given centroid is good (centroid that can represent the members in a cluster), then the clustering results can be ensured also good. Yet, it does not always happen, so it can be ensured that its cluster formation changeable (inconsistent). Centronit is an initialization method development of the initial centroids in K-Means. The algorithm of this method is based on the calculation of the average distance of the nearest data use the minimum distance. This method also robust from outliers of data. By using centronit method, the initial centroids can be determined in advance so it can produce the better and consistent cluster formation

Kata Kunci : N