<![CDATA[The k-means clustering method is a non-hierarchical grouping method that groups data into several centroid centers. The simplicity of the k-means method is widely used in various fields because it has several advantages, namely it is easy to implement and has a high level of accuracy of the size of the object so that this method is relatively more measurable and efficient. However, the initial k-means algorithm calculates using a C (centroid) value that randomly causes random results. Dependence on C (centroid) values ​​makes the accuracy of the k-means algorithm less than optimal. The results of k-means calculations are often obtained by experimenting several times and tend to produce different clusters. But in getting better results, it is difficult to determine the limits of an experiment. The random determination of cluster centers causes the k-means method has not been able to get the best grouping results. In this study, we describe an algorithm that is also used to optimize the selection of the initial center point in the k-means method, the pillar algorithm. This algorithm is an initial centroid determination by calculating the distance of metric accumulation between each data and all previous centroids. The choice of points is determined by data points that have a maximum distance. This research determines centroid using the Pillar algorithm and the results of the algorithm are used for the cluster's focal point on the k-means algorithm. In each cluster pillar algorithm is able to get the value of Sum of Squeared Error (SSE) better than random centroids as evidenced by the decreasing value of SSE.]]>
Copyrights © 2019
